JC_Emanuel

This is a long (20 page) and developed proposal. I've sent a separate email with the marked-up word document with several margin comments). Below are some bullet point comments that I included as well.

This is a very interesting and policy important research topic. Here are some broad comments on the written proposal and the questions it poses:

• Think about rewriting and organization. Right now you spend a very long time describing the problem and existing literature before getting to your own research hypotheses and design (which does not come until page 11). It would be more effective if you rewrote to produce a shorter (or at least reorganized) proposal. You should aim for a more succinct summary of the problem and your research questions on the first couple of pages. I suggest moving the 'literature review' section to the end (or keep it for later). I feel that in the first draft, you'd placed too much of the lit review up-front before you've made clear to the reader what your own approach will be.

• Even as the lit review was quite complete, I think you might improve it a bit. Table 1 was useful (though not before your main hypotheses) but I felt you could tie the different theories a bit more to the economics literature. Somewhat absent for example is any explicit reference to the Becker/Rosen human capital investment approach, namely the idea that individuals make the decision whether to continue an extra year of education by comparing marginal benefits (expected return to education) to the costs. This approach is implicit in some of what you lay out, but could be made more explicit. Individuals might then drop out because the perceived returns to educatin are very low (e.g. in a low quality school) or because the opportunity cost is high (e.g. good prospects for teen employment). One can fit some of your existing review around this... for example the several explanations where individuals cannot achieve an optimum of staying in school longer because of the types of shocks you mention. Other classic economics of education references might be added as well (e.g. Hanushek on the 'school production function' approach that you also reference.)

• You've put some emphasis on school/teacher quality, but perhaps not enough and you don't say much at all about direct measurement. There are some such as Lant Pritchett (see his CGD book Rebirth or Education available via GC library, or his blog posts) who make the point that the 'quantity' of education has risen very quickly around the world (new schools built, new compulsory education laws, etc) but that quality remained poor.. so and this explains dropout rates. His argument in a nutshell is that many students who enter middle school or high-school are just not prepared (as measured say by PISA scores) but they are moved forward through a curriculum that they don't understand or like, so they drop out. (Aside: Pritchett https://lantpritchett.org/education/, goes much further in his ascerbic critiques arguing that schools in many countries, particularly poor areas, are bad and unaccountable and the teachers badly trained and often abusive (if they even bother to show up for work) I doubt the situation is so bad in Argentina but he also makes the point that economics researchers tend to be, by revealed preference, people who liked school and therefore see dropping out of school as a failure and have little empathy or understanding of just unpleasant the experience of going to schools is for many students).

I mention all this to ask you if you have any direct measures of school quality, for example standardized test scores. You mention spatial variations in quality but don't really mention how you might mention school quality, which seems critical. Next-door Chile shares standardized test score data with researchers, leading to may education papers. I don't know if something similar exists for Argentina.

• Your list of hypotheses focuses on household variables, shocks, and 'spatial clusters'. That last category is a bit unclear. Are you referring to socioeconomic status, to provincial policies or insitutional quality. Again, can you measure any of this directly?

A few more minor things (no need to respond.. just things to think):

• clarify the relation between 'overage' and 'in-grade retention'.

• You mention some policies such as the 2006 extension of compulsory schooling that might potentially act as natural experiments (though maybe not because enforcement might be weak and there are less clear age cutoffs as in US mandatory schooling rules).

• what identification challenges? How to argue that the model on p14 is recursive? Has this model been used to study this question in other contexts?

• You write: "the study will exclude individuals that are head of household. The household head is the person who earns the greater proportion of the total income of the unit." Would this select out individuals who drop out of school because a shock pushes them to become head of household, or who formed new households because of teen pregnancy? Does the data allow any other definition of head of household?

• Data sources: Can you find test score data by province, district, or even school? You might then use average test scores say at 6th grade to measure quality directly.