EDUCATIONAL OUTCOMES IN OECD COUNTRIES
15
be preferable, these should still capture the most basic overall patterns, under the
assumption that main institutional variation is in the cross-section. At least, results are
unaffected by using the available indicators for 2003 or 2008 instead, or by taking the
average over the three observations, indicating that lack of results is not driven by single
measurement error. In addition, to align the regulatory measures more closely with the
period of growth observations, we performed all regressions for growth between 1980
and 2000. Results were unaffected.
The final five columns of Table 4 employ aggregate and sub-indicators of labor market
regulation, which again range from 0 to 6 with higher values reflecting more restrictions.
Columns (7) and (8) add the two versions of the aggregate employment protection index
suggested by the OECD to the model. The first version combines regulations of regular
employment contracts and of temporary contracts, and the second version adds sub-
indices of additional regulation of collective dismissal to this. Neither measure enters
the model significantly or affects the estimate on cognitive skills. The same is true when
the three sub-indicators of protection of permanent workers against (individual)
dismissal, strictness of regulation on temporary forms of employment, and specific
requirements for collective dismissal separately (columns (9)-(11)).
The indicators of employment regulation are measured as averages of the annual
values between 1985 and 2000. Results are similar when the growth period is restricted
to 1980-2000, which aligns more closely to the period of observation of the regulatory
measures, is used instead of 1960 to 2000 (not shown). Results are also robust to the
new, third version of the aggregate OECD employment protection index, available only
in 2008, which adds the maximum time to make a claim of unfair dismissal,
authorization and reporting requirements for temporary work agencies, and regulations
requiring equal treatment of regular and agency workers at the user firm as three new
sub-indicators.
Beyond the regulatory measures reported in Table 4, Nicoletti and Pryor (2006) survey
a total of 16 objective and subjective measures of governmental regulation from four
different studies, including Botero et al. (2004), Kaufmann, Kraay, and Zoido-Lobaton
(1999), Pryor (2002), and earlier versions of the indicators used above. While the above
objective indicators are based on extensive reviews and quantifications of laws and
regulations, other more subjective indicators are based on observational data of experts
familiar with the regulations. The different indicators also cover different sets of
regulations on product and labor markets, as well as overall business regulations more
generally. Not a single of the regulation measures is significantly associated with long-
run growth across OECD countries in our basic model (not shown). These results
confirm that the lack of a significant association with long-run OECD growth does not
hinge on a specific objective or subjective method of quantifying institutional realities or
on a specific area of regulations.
The results of including institutional and regulatory measures in the growth regressions
rule out that the strong association of educational outcomes with growth is just driven by
omitted causal institutional factors. In fact, the analysis provides no convincing
evidence that institutional or regulatory differences can account for differences in long-