We need to decide on a research question for Wed draft. I'm listing some that I thought of; open to any of these and any you guys have.
There is a different paper using the same data that takes transfers as exogenous changes in village inequality. They use this to measure the effect of inequality on life satisfaction but we can change the depvar with the same specification.
Relatedly, we can use the random transfers to men/women as exogenous change in wealth inequality by gender and look at the effect on gender based violence, labor market outcomes, etc.
Use random survey timing to estimate an impulse response function for the intervention on the standard outcomes.
The original paper has a selection problem wherein the control households were sampled later than the rest. This is a problem because the eligibility criteria (thatched roof) means they might have excluded people who upgraded to metal roofs during that time. We can do some correction (Heckman, PSM, synthetic control, etc.) and check whether the original results are robust.
We need to decide on a research question for Wed draft. I'm listing some that I thought of; open to any of these and any you guys have.
There is a different paper using the same data that takes transfers as exogenous changes in village inequality. They use this to measure the effect of inequality on life satisfaction but we can change the depvar with the same specification.
Relatedly, we can use the random transfers to men/women as exogenous change in wealth inequality by gender and look at the effect on gender based violence, labor market outcomes, etc.
Use random survey timing to estimate an impulse response function for the intervention on the standard outcomes.
The original paper has a selection problem wherein the control households were sampled later than the rest. This is a problem because the eligibility criteria (thatched roof) means they might have excluded people who upgraded to metal roofs during that time. We can do some correction (Heckman, PSM, synthetic control, etc.) and check whether the original results are robust.