Education is one of those subjects where evaluation of results is most difficult. Not only is it difficult to separate value added from one year of schooling from already existent human capital (or ability), but also due to the fact that most of the effects of education are felt in the long-run. Moreover, when one wants to compare policy, one must also be able to compare results in one school with another, in different years and between countries if possible. Finally, different policies can yield similar gross results albeit with very different cost effectiveness.
One of the methods created to address some of these concerns was the Programme for International Student Assessment (PISA). It was held first in 2000 and currently it evaluates Mathematics, Science and Reading in 65 countries, every three years. This was a huge advance, for it allowed to compare results in different countries, which naturally have different curricula. Also, being somewhat similar in difficulty, it should also be comparable across different years.
However, some caution is needed when doing analysis on this kind of data. One must always remember that this is aggregate data. One of the problems with aggregate data is that, by looking at it, one might interpret relationships at an aggregate level that do not actually hold at the individual level. This is called an ecological fallacy. In the case of education, we are coupling together different schools, students in a country average that might not make sense. What is more, PISA reports look mostly at correlations, but they cannot infer causality from cross-country data. Finally, there is another problem, which in my view is the most important. Policies in education take time to act, and thus it is not easy to assign results to given policy.
Nevertheless, PISA results (and others, like TIMMS) are still a significant step in gathering better data for accurate analysis. Also, there are indeed some papers like Ramos et al. (2012) do use PISA microdata to make analysis and draw informed conclusions.
José Miguel Cerdeira