Introduction
Although there has been tremendous progress in expanding school enrollments and increasing years of schooling in recent decades, 113 million children of primary-school age are still not enrolled in school (UNDP, 2003).
This paper reviews what has been learned from randomized evaluations of educational programs about how best to increase school participation. I first outline the intuition behind the important role of randomized evaluations in obtaining credible estimates of the impact of educational interventions, review other non-experimental methods of evaluation, and present some evidence on the biases that can arise with such non-experimental methods. I then discuss two types of programs that have been found to be extremely cost-effective and that could be implemented even within a very limited budget: school-based health programs and remedial education programs (that take advantage of available inexpensive sources of labor). I then outline a series of programs aimed at lowering the costs of school, or even paying students for attending school, that could be implemented if a higher level of financial support is available and discuss the possibility of educational reform through school choice, which could be implemented given sufficient political will within a country. The paper concludes by drawing some general lessons about the contribution of randomized evaluations to understanding the cost-effectiveness of various interventions.
Given the widespread consensus on the importance of education and the several reviews of the impact of education on income and other outcomes, this paper focuses not on the effects of education but on issues internal to the education system. The scope of this paper is also limited in that it does not address interventions intended to improve the quality of education, such as computer-aided instruction, unless these interventions cut costs and thus free resources that can be used to expand education.
