More countries, similar results. A nonlinear programming approach to normalising test scores needed for growth regressions

Analysts such as Hanushek and Woessman have brought to the fore the deceptiveness of education enrolments, or years of schooling, in growth regressions and the need to consider educational quality. In this paper, a nonlinear programming solution is proposed as a way of normalising to a single scale country average test scores from various international testing programmes. This method, though less transparent and more dependent on certain subjective choices than the existing approach put forward by Hanushek and Woessman, allows for the inclusion of more countries, in particular more African and developing countries, into a growth regression. The regression produces the results one would expect, namely a strong conditional correlation between growth and educational quality. The utility of growth regressions with an educational quality variable for the education policymaker is discussed. A method for arriving at feasible annual improvements in educational quality and hence feasible country targets is presented

C14, I28, O15

human capital, cross-country growth model, test score data, nonlinear programming, education policy, PISA, SACMEQ, SERCE