Causal inference

[Ishwor Adhikari]

Survey Panel Data of 6000 households, with 400 Primary Sample Units(PSU) wave 1, wave 2 and wave 3. Some PSU receive seasonal labor while other does not. Receiving seasonal labour is not systematic in the sense that
Some PSU receives seasonal labor only in wave 1, around 21 PSU, some receive only in wave 2, around 16 PSU, and 88 PSU receive seasonal labor only in wave 3. Rest of the receiving PSU receive in mix of the waves .i.e 16 PSU receive in wave 1 and 3, 10 PSU receive in wave 1 and 2, 56 PSU receive in wave 2 and 3 while 27 PSU receive in all waves. My outcome variable is at the household level say expenditure on hired labour, and the interest variable is receiving seasonal labour which is at PSU level. Unbalanced Panel. Can I establish a causal relationship between the two? What would be the best strategy if it is possible? The data comes from Survey Data not RCT, meaning seasonal labor receiving is characteristic of the community we observed not a policy intervention.

[Jared Greathouse]

I started reading at

My outcome variable is at the household level say expenditure on hired labour, and the interest variable is receiving seasonal labour which is at PSU level

My question for you, is does the treatment turn on once and stay on for every treated unit? (i.e, units aren't treated twice)?

[Ishwor Adhikari]

Does the treatment turn on once and stay on for every treated unit? (i.e, units aren't treated twice)?

I guess you are calling receiving seasonal labor as 'treatment'. The answer is No. Treatment does not stay on always. For some, it stays for some it does not. I tried to explain it above stating

Some PSU receives seasonal labor only in wave 1, around 21 PSU,.......

[Maxence Morlet]

I recommend you give this paper a read: https://papers.ssrn.com/sol3/papers....act_id=3906345.

[Ishwor Adhikari]

Maxence Morlet I did went through

Two-Way Fixed Effects, the Two-Way Mundlak Regression, and Difference-in-Differences Estimators by Jeffrey M. Wooldridge :: SSRN

. But I could not find my answer. Jeff Wooldridge do you like to comment on this issue? Or anyone in this forum? Does DID staggered exit work in my case?

[Ishwor Adhikari]

Does anyone like to comment?

[Ishwor Adhikari]

193 PSU never received seasonal labour.

[Jared Greathouse]

Staggered DD works here. The paper Max recommends is good reading. Another paper you read should be this one.

[Ishwor Adhikari]

Thank you Jared Greathouse I will go through it.

[Ishwor Adhikari]

I am unable to find literature that could deal with the problem I am facing. Any suggestions?

[Jared Greathouse]

The paper i recommend provides updates on the newest methods in DD which were designed specifically for this problem. Literally, it discusses staggered adoption in the paper, so what issue are you facing that the newest estimators can't adjust for?

[Ishwor Adhikari]

The issue is how to do away with the assumption that treatment remains once treatment starts. In my case, some (PSUs) get treatment in wave 2 and don't get it in wave 3. How to take care of it? Jared Greathouse Do i need to drop observations (i.e. PSUs) who got treatment in the first wave itself? What to do with observations of who got treatment in all three waves? Do I need to drop them off also?

[Jared Greathouse]

Look up the flexpaneldid command

[Ishwor Adhikari]

Thank you Jared Greathouse for your continuous input. I think I should look into some other alternative methods than to DID framework as it does not seems to apply in my case, given there is no pre-treatment status for many PSUs. I am thinking of modelling something like this

y_it x_pt PSU time , where i is household, x is dummy whether PSU has seasonal labour(=1) or not(=0) at time t, PSU is PSU fixed effect, and time is fixed effect.

What do you suggest?