Running a Tobit with panel data

[Arpan Sagar]

Hey all!

As a part of my Master's thesis, I am conducting an analysis of underreporting of migration statistics in Eastern Europe. For this purpose I have computed an Implicit Minimal Measurement Error (IMME) (Link attached) which takes a value between -1 and 1. I plan on estimating a causal relationship between the IMME and a few key migration drivers (gravity equivalents of economic, demographic and social factors). Initially planned on running a simple panel regression with country pair (panel variable) and year fixed effects, however, since, my dependent variable takes a value only between -1 and 1, I feel like I must opt for a MLE in the form of Tobit. Could you please help me identify the command for running a panel Tobit for my study Please note that I am working with an unbalanced panel as countries have differing reporting periods as well as migrant compositions





https://repub.eur.nl/pub/100849/

[Carlo Lazzaro]

Arpan:
welcome to this forum,
See -help xttobit-.

[Arpan Sagar]

Hi Prof Carlo

Thanks! Have been running my model with xttobit itself, however, I wonder what the tradeoff is due to the lack of inclusion of country pair fixed effects in the model and if there exists an alternative? Would you consider pooled OLS with clustered standard errors to be a viable alternative (As it also allows me to include time invariant variables such as distance and borders)

[Carlo Lazzaro]

Arpan:
most depends on how much a pooled OLS is acceptable for dealing with this kind of issue in your research field.
In addition, please call me Carlo, like all those on (and many more off) this list do. Thanks.

[vikas mishra]

Hello stata users,

I am running xttobit random effect on balance panel data. After that I want to estimate the marginal effect.

However, i am getting the same estimates from marginal effect step similar to xttobit regression

xttobit equation :
xttobit quantity ln_price ln_price2 household_size ln_household_income, ll (0.0) tobit intpoints(20)

y is censored below at zero.


marginal effect equation :
margins, dydx(*) predict (ystar(.,.))

Is there any specific mistake that causes the similar estimate in marginal effect calculation.