EDUCATIONAL OUTCOMES IN OECD COUNTRIES
standard deviation of 100 across OECD countries, and then all available test scores are
aggregated to the country level.
We interpret the test scores as an index of the human capital of the populations (and
workforce) of each country. This interpretation of our averages over different cohorts is
reasonable if a country’s scores have been stable over a long period, implying that
estimates from the current school-aged population provide an estimate of the older
working population. If scores (and skills) do in fact change over time, some
measurement error is clearly introduced. We know that scores have changed some
(Hanushek and Woessmann (2009)), but within our period of observations it still appears
that the differences in levels dominate any intertemporal score changes.8 Nonetheless,
any measurement error in this case will tend to bias downward the estimates of the
impact of cognitive skills on growth, so that our estimates of economic implications will
be conservative.
The data on GDP per capita and its growth for our analyses come from the Penn World
Tables (Heston, Summers, and Aten (2002)). Data on quantitative educational
attainment are taken from the latest version of the Barro and Lee (2010) database.
Additional measures of specific control variables will be discussed in the specific
sections below.
Table 1 provides basic descriptive statistics on the combined measure of educational
performance and the underlying economic data. We have already discussed the wide
variation in growth rates across OECD countries. What is also clear from Table 1 is that
both school attainment and test scores vary widely, suggesting directly that any impact
of these human capital measures on growth differences should be easily detected.
2.3. Basic results for OECD countries
We use the data on educational outcomes to estimate cross-country regressions that
describe long-run growth for OECD countries. This follows a growing literature which,
over the past ten years, demonstrates that consideration of cognitive skills dramatically
alters the assessment of the role of education and knowledge in the process of economic
development. Analyzing growth in 1960-1990 for a sample of 31 countries with
available data (including 18 OECD countries), Hanushek and Kimko (2000) first showed
a statistically and economically significant positive relationship between cognitive skills
and economic growth. This relationship between cognitive skills and economic growth
has been subsequently confirmed in a range of studies with different focuses.9 Most
recently, Hanushek and Woessmann (2009) extend the empirical analysis to incorporate
50 countries that have participated in one or more international testing occasions
between 1964 and 2003 and have aggregate economic data for the period 1960-2000.
We use that database for our analysis focused on OECD countries.
88 The subsequent simulations that investigate the impact of changing achievement for each country assume that the increments
to achievement change for each cohort affects the labor force skills proportionally to their weight in the age distribution.
9 Important studies include, for example, Barro (2001), Woessmann (2003b), Bosworth and Collins (2003), and Ciccone and
Papaioannou (2009). See Hanushek and Woessmann (2008, 2010) for reviews.