Hi there,
I have to estimate the causal effect of part-time work compared to full-time work on health. I have an unbalanced panel data set with t=2008-2017, except for year 2009.
The main issue I have is how I should deal with reverse causality. Every article I read explaining this uses an IV with FE. However, I believe I don't have a good IV in my data set. Is there another way to deal with this? I'm coming across lagged variables, but I have no idea how I should implement that and if that would deal with reverse causality. Could you maybe help me?
Thank you in advance.
Variables in my dataset (in case someone has a good idea for an IV)
- person id number
- age
- number of kids
- civil status (i.e. married, divorced, etc.)
- home owner or rental
- urbanity of place of residence (closer to city/rural area)
- education level
- self-assessed health
- disability (suffer from long-standing disease, affliction or handicap, consequences of acccident yes or no)
- hinder (to what extent health hinder individual's work)
- gender
- survey year
- employed (paid employment)
- job hours (weekly hours paid employment)
- part time (1 = 32 hours or less, 0 = 33 hours or more, 0 hours is missing value)
This paper proposes an idea based on the panel structure:
https://www.diw.de/documents/publikationen/73/diw_01.c.466650.de/diw_sp0663.pdf
Are there any variables related to the employment status of peers? One can argue that if a larger share of individuals in the peer group are full time employees it may increase the likelihood of individual i being a full-time employee as well; however, theoretically share of full-time employees in the peer group may not affect individual's health. This can be considered as an instrument in the first stage along with other covariates and FE.
Shreya Biswas
Thank you for your response! Unfortunately there is no data about that in the set we are allowed to use. I thought maybe I have to do something with the variable that indicates to what extent someone's health may hinder their work, since that might say something about whether there is reverse causality. However, I don't know if I'm going in the right direction and if so how I should use that.
- Badges
- Science topic