We construct an unbalanced panel that utilizes data on comparative international tests in
Mathematics and Science for the age group 13-15 years and includes eight tests in the period
1980-2003 for a maximum of 79 countries. This panel allows the application of panel data
estimation methods, considerably improving the methodology of previous analyses in
education economics. Existing studies utilizing cross-country variation in student achievement
are either almost exclusively based on a cross-section of individual test performances in the
same test of a single year or the country average in performance across many years as in
Hanushek and Kimko (2000), Hanushek and Wossmann (2007), and Jamison et al. (2007).3
We follow to some extent the latter methodological approach, but exploit the panel structure
in the data and estimate country fixed effects models that account for unobserved
heterogeneity across countries.
The empirical analysis is preceded by a simple theoretical model that relates the size and the
scope of the welfare state to human capital investment in terms of student effort. The
insurance aspect of the welfare state manifests in a system that both reduces the risk of future
income and that redistributes from high income individuals to low income individuals. In
traditional human capital models (Becker, 1964), where educational outcomes are determined
by rational individuals weighting costs and benefits, increased redistribution of income is
predicted to weaken the incentive to invest in education. The prediction of the effect of
reduced idiosyncratic risk in future income is more complicated and ambiguous.
The paper proceeds as follows. The next section presents the theoretical considerations.
Section 3 describes the international student tests data and our measure of adjusted average
student performance, section 4 presents the empirical model, and section 5 discusses the
empirical results. Section 6 offers some concluding comments.
2. Theoretical considerations
We present a simple partial two-period model that illustrates how redistribution of income and
income uncertainty may affect students’ incentives in schools. The model builds on Glomm
and Ravikumar (1992), but differs by including income redistribution and uncertainty.4 At the
outset we consider a model with one decision-maker, the student, but below we discuss its
real-world applicability where other agents such as parents and teachers have an additional
influence on student effort.
3 The only exception seems to be Barro and Lee (2001a), who employ a panel of countries participating in
international tests up to 1990 to estimate the effect of school inputs on student achievement.
4 Another related theoretical paper is Poutvaara (2007) who relates the old-age pension system and migration
prospects to investment in human capital.