empirically identify mechanisms through which governments’ welfare and redistributive
policies may affect aggregated student achievement.
From our model, the implicit function of education investment is
Z = f (τ, σ12,y ) = Z ( g,y )
where g=g(τ,σ12) captures the common effects of the welfare state. The investment Z
decreases in y because the utility function is concave, while the effect of the welfare state g is
in general ambiguous. However, redistribution of income, τ, is regarded as the dominating
aspect of the welfare state and has a negative effect on Z. Our testable hypothesis is that a
more generous welfare state affects individuals’ educational investments negatively.
3. International measures of student achievement
We rely on comparative international tests of student achievement conducted by different
international organizations. The International Association for the Evaluation of Educational
Achievement (IEA) has been responsible for the largest number of such tests, among them the
TIMSS tests, but also the OECD has developed a Programme for International Student
Assessment (PISA).
We construct a synthetic panel data set of the national averages of international tests on
student performance covering a period of almost 25 years (1980 - 2003). These tests cover the
core subjects Reading, Mathematics and Natural Science separately, but we restrict our
attention to student assessments in Mathematics and Science for several reasons. First, these
two subjects have more similarities with each other than with Reading and are thus more
suitable for constructing a synthetic panel. Second, reading skills are tested less regularly
internationally, and even within the same test and year, potentially differ considerably by
language. Third, performance in Mathematics and Natural Science are more likely to
determine a country’s innovativeness in an economic growth context, as empirically tested in
Hanushek and Wossmann (2007). Comparability of test results is also given in the age
dimension, as all tests included are conducted on middle-aged students (13-15 years).
Choosing this age group has also the advantage that compulsory schooling still applies,
mitigating selection out of education issues. The tests we utilize are described in Table 1.