EDUCATIONAL OUTCOMES IN OECD COUNTRIES
20
different scenarios of school improvements in the OECD member states. The
projections presume that the estimated impacts of cognitive skills on growth are causal
in nature - that is, changing the achievement of a country’s population will lead to
improved growth. Of course, considerable controversy surrounds such cross-country
growth regressions and any causal interpretation of them (see, for example, Levine and
Renelt (1992), Levine and Zervos (1993), and Bils and Klenow (2000)). In other work,
we have considered a series of analyses aimed at eliminating many of the natural
concerns about the identification of the causal impacts of cognitive skills (Hanushek and
Woessmann (2009)). Each of the analyses points to the plausibility of a causal
interpretation of the basic models. Nonetheless, with our limited international variations,
it is difficult to demonstrate identification conclusively. Therefore, here we consider a
variety of sensitivity analyses designed to show the impact on outcomes when part of the
estimated effects is non-causal.
We consider three specific reform scenarios: first, improving average student
performance 1/4 standard deviation, or 25 PISA points, in each country; second,
bringing all OECD countries up to the level of the PISA top-performer, Finland; and
third, bringing all students in OECD countries to minimum proficiency, defined as 400
points on the PISA test scale. For each scenario, we aim to project the future path of
development of GDP in a country to calculate the economic value of the education
reform in present-value terms.
One final issue of the projections is important. Because we consider economic
outcomes far into the future, the precise form of the underlying growth model potentially
makes a noticeable difference. In particular, an endogenous growth model that implies
constant growth effects from improved skills will differ from a neoclassical growth
model that sees changes in skills as affecting the level of income but not the long-run
growth rate of an economy. Our previously estimated models support analysis of this
range of underlying models, and applying the alternatives gives another idea of the
bounds on future economic effects from school improvement.
5.1. The projection model
The projection of the total value of any education reform involves several components.
First, we calculate the time path of the annual growth rate engendered by education
reform designed to move students from their current performance to a given new level.
This pattern of economic outcomes represents the confluence of three separate dynamic
processes: (1) Changes in schools lead to the progressive improvement in student
achievement until students fully reach the new steady-state level of achievement; (2)
students with better skills move into the labor force and the average skills of workers
increase as new, higher achieving workers replace retiring workers; and (3) the
economy responds to the progressive improvement of the average skill level of the
workforce. Second, based on the pattern of predicted growth rates, we model the future
development of GDP with and without the education reform. Third, based on these