EDUCATIONAL OUTCOMES IN OECD COUNTRIES
30
horizon of projections is limited. Across the three reform scenarios, the neoclassical
projection value is between 65 and 83 percent of the endogenous-growth projection
value.
The difference between the two growth models depends on the time horizon chosen.
The further out the projections are carried, the more substantial does the difference
between the two growth models get. The final four columns of Table 8 report the
present value calculated with the estimated neoclassical version for reform Scenario II
with time horizons varying between 2050 and 2150, respectively. The varying time
horizons clearly make a huge difference in the estimated returns to reform and illustrate
how the neoclassical and endogenous growth specifications differ. Compared to the
endogenous-growth projections evaluated over the same time horizons (discussed in
Table 9 below), the neoclassical projection is 81 percent of the endogenous-growth
projection value for a time horizon until 2050, 65 percent until 2090, and 44 percent
until 2150.
Several factors contribute to the closeness of the estimates over our time period for the
impact of improvements in cognitive skills based on the two different growth models.
First, our reform scenarios gradually introduce changes, due to the lags for the policy to
become fully effective and for the new, better-educated workers to change the average
skills of the labor force. Our projections from the time after policies are fully felt
involve just 20 years, a time too short to have huge differences in the implications of the
alternative models. Second, the biggest impacts of the differences across the alternative
models occur in the distant future, and thus the impact is lessened by discounting to
obtain present values and by disregarding any returns that might accrue after 2090.
Third, even ignoring discounting, the estimated convergence parameters imply very long
periods before any country returns to its balanced growth path following a perturbation
because of policy. Fourth, the present value of any reform keeps increasing even after
full convergence has taken place and after growth returns to its pre-reform rates, because
the economy is at a higher level due to the reform.
As a consequence, for the (already long) time horizon taken here, the different
dynamics of the neoclassical and endogenous growth model do not make a fundamental
difference, regardless of whether human capital increases the long-run growth rate or
not. The alternative set of neoclassical projections places a lower bound on the results
based on the endogenous growth paradigm but retains the same order of magnitude.
5.4. Sensitivity to alternative parameter choices
The different scenarios indicate very large economic changes from educational
improvement, but the estimates come for specific parametric choices. How sensitive are
the results to these specific choices? We look at sensitivity in terms of reform Scenario
II of bringing each country to the Finish level of PISA scores in our baseline
endogenous-growth-type model specification. The qualitative implications for changes
in results, however, apply similarly for the other scenarios and growth model.