Income process for whom and one estimation issue

[zhangxiang0822]

For Luxi Han just a very quick questions.

• Income process for whom? The individuals or households?
• Also, I'm wondering if the endogenous labor supply of household members will affect the income process estimation since the self-insurance of households will attenuate the degree of permanent shock. In this way, if it's impossible to disentangle the permanent shock to household response thus impossible to use regression to estimate income process. (see Blundell et al. 2016)

[magicahan]

Hi! It's the income process for individuals. There's some room left for risk sharing. I think this has been discussed in Kaplan's paper. But I'm not sure if I'll have room to discuss that.

Yes, definitely. But now with the income process estimated, the simulation exercise assumes an exogenous income process given by the estimation. Then adding labor supply would need other modification. But potentially, the labor supply decision is embedded in the transitory component. I'll need to think about this more. But the extensive margin can potentially be more interesting on my end, since we have observations for unemployment spell and earnings. Then I think we could definitely come up with a model with indivisible labor supply.

[magicahan]

One way to think about this. Log(earnings) = Log(w.L) = log(w) + log(L). Then the permanent and transitory component will be embedded in this problem. The way I think of it, is that the response in log(L) can be in the transitory component (people's labor supply won't change permanently by standard business cycle assumption). This puts the labor supply endogenous in the income process.

This would even more justify our assumption of correlation between permanent and transitory component. If the response is endogenous, then the independence assumption may be violated. But in our approach, we are just interested in the estimation of the nonparametric distribution. But you are right, the labor supply issue is tricky. But our data has hours worked for employed people. I can see if I can do something with it. Thanks for the advice!

[zhangxiang0822]

Oh yeah, the household labor supply decision could be a transitory response (while labor supply shock to permanent shock could be long-lasting).

But individually, I think if the income process you're estimating is at the individual level, it might won't be necessary to incorporate household risk-sharing decisions to your model(I'm not familiar with the literature in this field).

Or if you want and you have data (like you can linking spouses, I think PSID is a good data source), possibly you can do something with that. But I'm not sure whether this will make you divert from your original path since it might not be an important issue at all! :)