In a pinch, extrapolating the short-term effects of these programs is not necessarily a bad idea, particularly if one assumes that their effects persist over many years. Yet research by Andrabi, Kas, Khwaja, and Zajonc (2011, hereafter AKKZ) suggests that the treatment effects of educational …show more content…
Measuring this effect precisely is important – the benefits of educational attainment on a country’s productivity are well established, and in many developing countries, private schools cost less per pupil than their public school counterparts do. (AKKZ use a dataset on elementary school students from Pakistan, but the effects of private schooling in Pakistan may mirror those in other developing countries as well. ) If private schools can improve educational outcomes while costing less, perhaps governments should encourage their growth through school voucher programs. Unfortunately, though, measuring the long-term effects of any educational program is far from easy, and requires the economist to overcome three key issues: learning persistence, unobserved heterogeneity, and measurement error. Before examining these issues, however, an introduction of AKKZ’s model of educational achievement is in …show more content…
with the alternative model being a normal linear regression) returns a χ^2-value of 533.72, which corresponds to a p-value of 0.0000. The model clearly suggests that learning persistence differs among students. More importantly, however, learning persistence also differs considerably among students – indeed, with an estimated value of approximately 0.25, the standard deviation of ζ is quite sizable compared to the value of β.
CONCLUSION
Even if their model is fundamentally misidentified, as suggested by Hansen’s J-test, AKKZ’s paper provides some valuable contributions to the literature. Indeed, they clearly show the importance of including persistence in estimating the long-term effects of educational programs such as private schooling. In addition, if AKKZ’s estimates of the effects of private schooling are accurate, governments in developing countries should encourage the growth of private and charter schools, which are both cheaper per pupil and potentially more effective in raising test scores.
Going forward, however, researchers would be wise to remember that educational programs are rarely “one-size-fits-all,” and should consider random coefficient models in future papers. Equally as importantly, researchers should closely examine models of forgetting to ensure that learning persistence is truly geometric – particularly in educational environments where teachers consistently review previously taught