Page 1 of 1

a few issues

Posted: Tue Jun 12, 2018 5:44 am
by Luca Tiberti
Thanks for the great presentation and congratulations for the impressive improvements since your first submission of the proposal. I have a few comments:

- I am not convinced about the validity of the instruments, both in terms of exogeneity and in terms of the exclusion restriction
- also, note that your "working while studying" (WS) variable (but also your outcomes) is observed over a large time span, in most of the observations BEFORE 2015 (the year of your poverty variable). So, it seems to me that the instrument is weak
- it is important to define clearly the WS variable. In particular, it can include working during the school year (the whole year), occasionally during the academic year or during summer. The impact mechanisms can be completely different depending on the definition. If there are enough of observations in each of the different options, you should separate the definitions (at least in robustness analyses)
- you briefly talked about sample selection, but it is unclear which is the source of it, and how you propose to address it.



Posted: Tue Jun 12, 2018 7:08 am
A possible concern: the number of children, proposed as an instrument for internship, may potentially impact the outcome variable (mismatch, over, under) directly if employers are have, say, preferences against hiring women with a high number of children. there is a huge literature that looks at the impact of fertility on labour market outcomes, treating fertility as endogenous.