EDUCATIONAL OUTCOMES IN OECD COUNTRIES
18
but the point estimate on the top-skill dimension is substantially higher. A ten
percentage-point increase in the basic-skill share is associated with 0.3 percentage points
higher annual growth, and a ten percentage-point increase in the top-skill share is
associated with 1.3 percentage points higher annual growth. (Note that this does not
necessarily provide an estimate of the relative importance of the two skill dimensions, as
it may be much more feasible to increase the basic share than to increase the top share by
the same amount; this might be suggested by the fact that the international standard
deviation of the basic skill percentage is about four times as large as that of the top-skill
level).
When estimating the same model on the OECD sample, though, the point estimate on
the top-skill share is only a fourth of the one estimated in the full country sample and
loses statistical significance (column (2)). By contrast, the point estimate on the basic-
skill share is slightly larger than in the full country sample, and remains highly
significant. The specification of column (3) shows that the difference in the estimate on
the top-skill share between OECD and non-OECD countries is statistically significant.
Of course, the measures of the two skill dimensions are highly collinear (their correlation
is 0.73 in the full sample and 0.70 in the OECD sample), limiting precision in the joint
specification. Results in columns (4)-(9), however, reveal that the pattern of results is
similar when entering one of the two measures at a time.
While small samples and collinearity obviously suggest caution in the interpretation of
these results, it seems clear that basic skills are relevant for OECD-country growth.
Furthermore, if anything, high-level skills appear more important in non-OECD than in
OECD countries, rather than the other way around.
4.3. Non-tertiary vs. tertiary schooling
The available test score measures - for average, basic, and top skills - are measured at
the primary and secondary level of schooling. However, a lot of policy discussion
focuses on another dimension of schooling, namely whether developed countries should
place a particular focus on primary, secondary, or tertiary education. Vandenbussche,
Aghion, and Meghir (2006) suggest that countries close to the technological frontier
should rather emphasize tertiary education. To investigate this dimension, we make use
of the new, improved Barro and Lee (2010) database which provides average years of
schooling separately at the primary, secondary, and tertiary level. With little meaningful
variation in the completion of primary education across the bulk of OECD countries, we
combine the two basic levels of schooling into one category of non-tertiary schooling.
In the full-country sample, the coefficients on non-tertiary and tertiary schooling are
both close to zero when cognitive skills are controlled for (column (1) of Table 6). By
contrast, in the OECD sample, the point estimate on years of tertiary schooling becomes
larger (column (2)), and reaches marginal significance (at the 10 percent level) when
years of non-tertiary schooling are not included in the model (column (3)). However,
results in column (4) indicate that this is completely driven by the United States. Once