comments and suggestions

Post Reply
Luca Tiberti
Posts: 33
Joined: Thu Jun 07, 2018 6:31 am
Contact:

comments and suggestions

Post by Luca Tiberti » Fri May 31, 2019 4:08 pm

Dear Sim and team,

thanks for the presentation. Here below some comments and suggestions:

As suggested, try to get your paper edited by an English editor in your institute (no problem if this will not be perfect – PEP will make proofreading later on – but the current version is sometimes hard to follow)
I would improve the structure of the paper. I am sure that your mentor can guide you well on this. I am happy to discuss further on this if you like.
You state that the study “provides a new perspective on the setting of Cambodia”. In which terms is this new?
External validity: this is an issue, but I would at least try to support with evidence that your sample (and, thus, results) is not “too much” different with respect to all rural areas in Cambodia. You could compare some key variables estimated with this survey and with a nationally representative
PSM:
- control variables only include « demand-side » related variables. I would also include supply-side variables, such as availability of schooling in the child’s neighboring (distance to school?)
- No need to provide a so detailed description of the PSM estimator.
- Say clearly which bias you would correct with the “nnmatch” command
- Good that you used pre-treatment control variables. However, since we do not know when migration started, it seems hard to say that your control variables are “pre-treatment”. Please argue better.
- I would report, in appendix, the PSM estimates. Also, we do not know anything about the control variables (I just guess they are the same as in the descriptive stats)
Identification strategy: I liked it (together with the falsification test with 2017 data – but very few treated and controlled obs/is this an issue for the validity of such a test?). You could run the bounds method proposed by Rosenbaum (2002) to see the probability that unobservables affect your estimations, but this is still not fully satisfactory. However, with the data at hands, I would have probably adopted a diff-in-diff combined with PSM.
Data:
• 10% attrition rate: you should at least show some descriptive stats for households who stayed in the sample and those who were lost
• You constructed an individual panel (which is key to your analysis), but you should clearly say how you did it (i.e., which were/was your merging variable(s))
Definition of migration variables:
- some of the definitions look unclear (see number 2, but also the definition of “absent”);
- do not know whether they remit or not, whether short or long term?
- Which are the reasons for the migration (from the presentation I understood you included “job” and “study”, but these two reasons are expected to have different impacts on children) – please clarify
- Ideally, fathers and mothers separately as we expect that fathers’ and mothers’ migration affect child wellbeing differently
- I understand that you have too few obs for some of the definitions, but you should clearly say that.
Outcome (number of years of schooling): this is a cumulative indicator. Hence, it is important to define a clear timing of the impact pathway. But without knowing when the parents left, it is hard to define a clear pathway mechanism.
Results:
- Some counterintuitive figures in the descriptive stats (Table 3) (e.g., large difference in terms of sex, better father’s and mother’s education but treated households are poorer, etc);
- Please check the p-values of your results of the bias-adjusted, as there seem to be some mistakes (e.g., Table 5)
Heterogeneity analysis of your PSM results:
- Analysis by age: 10-12 are overlapping between the two analyses, why?
- Add a few others: by, e.g., propensity to migrate, amount of remittances, etc.
Impact pathways:
- Your main assumed impact pathway is the “added work” by children left-behind to replace adults’ departure. Why do not test it directly on the data?
- The heterogeneity analysis by the level of remittances would probably something about the eventual role of the “income effect” pathway.

Post Reply

Return to “PMMA-20209 - Migration, Remittances and Child Schooling in Rural Cambodia”

Who is online

Users browsing this forum: No registered users and 1 guest