EDUCATIONAL OUTCOMES IN OECD COUNTRIES
17
and a threshold of top skills (see Hanushek and Woessmann (2009) for details). Using
such distributional information allows us to look at different dimensions of the skill
distribution separately.
Conceptually, Vandenbussche, Aghion, and Meghir (2006) assume that the innovation
process is more intensive in high-skill labor than the imitation process. They then
present an endogenous growth model with innovation and imitation where high-skill
labor has a greater growth-enhancing effect for countries closer to the technological frontier,
whereas countries further from the technological frontier get greater value from what they
call “unskilled human capital”. While the untested underlying assumption seems
reasonable, there are also reasonable arguments to be made for an opposite assumption -
in which case the main prediction of the model would be turned around. In particular,
the innovation literature points out that many innovations emerge from lucky
coincidences. By contrast, almost by definition purposeful imitation processes require
the employment of skilled scientists.
Another conceptual extension starts from the perspective of a high-skilled scientist. If
this scientist were to work in a country that produces at the technological frontier, his
only option is to use his skill in the innovation of new technologies. If the scientist, by
contrast, were to work in a country that produces far below the technological frontier, he
also has the option to employ his skills in such innovative activities, but he also has the
alternative additional option to employ his skills in imitating the more productive
technologies currently employed at the technological frontier. This scientist will tend to
choose to employ his skills in the activity that promises the higher benefits. In such a
setting, the return to high-skill labor cannot be smaller below the technological frontier
than at it, and it may well be larger. While concentrations of high-skilled labor and
spillovers across them may still be important, the alternative perspectives do introduce
questions about the underlying assumptions.
Given the conflicting conceptual predictions, the issue warrants a new empirical look.
In particular, while the empirical application in Vandenbussche, Aghion, and Meghir
(2006) measures growth in 5-year periods, which may be substantially influenced by
business cycles and idiosyncratic shocks across 19 OECD countries, we keep our focus
on differences in long-term growth experiences across the OECD countries. In addition,
while Vandenbussche, Aghion, and Meghir (2006) conceptualize the difference between
high- and low-skill labor as school attainment at the tertiary vs. non-tertiary level, our
alternative focus on top vs. basic skills may provide an empirical distinction particularly
relevant for imitation and innovation processes.
We start by replicating the results in Hanushek and Woessmann (2009) of
incorporating both the share of students who reach a basic level of skills (a score of at
least 400 on the tests) and the share of students who reach top-level skills (a score of at
least 600) in a growth regression of the full 50-country sample of OECD and non-OECD
countries (column (1) of Table 5).14 Both skill dimensions enter the model significantly,
14 These scores are one standard deviation below and above the OECD mean, respectively. The OECD also identifies five
levels of skills on each of its tests (see, for example, Organisation for Economic Co-operation and Development (2004)). The
400 falls in the range of level 1 while the 600 falls in the range of level 5.