I think I want to use some extensions of -ppml-, but my data is complicated and I’m unsure how to proceed. I am using Stata16 SE for Windows, for your information (I'm new to Statalist, and I've tried my best to follow the instructions on writing posts, but please let me know if I could do better!)
I want to estimate a model with the following data features:
- corner solution: the dependent variable is a production cost-share variable, bounded between 0 and 1, and has a lot of zeros. The zeros are present because some it's optimal for some fields to not use the input;
- multi-level fixed effect with a great number of dummies: the dataset is at the field level, spanning 2010-2016, and contains more than 75,000 field-level observations and around 14,000 farms. Note that it is not a panel, because each farm has several fields, and we can't identify whether the fields are the same fields across years. However, I want to control for the farm-level and state-level fixed effects (year fixed effect too, of course);
- dummy endogenous variables: the independent variables include two endogenous variables, and they are both dummy variables. From my experience with Tobit, dummy endogenous variables are trickier than continuous ones.
- to confirm my understanding is correct, that is, the PPML estimator is the most suitable for my problem, or at least a better option than Tobit. I've read some relevant threads before saying that, and particularly this one: https://www.researchgate.net/post/ST...nreyro_setting where Professor Santos Silva has expressed concerns about incidental parameter problem. My dataset has 7 years, but I'm not sure if it works.
- the right way to proceed. I have looked at extension commands such as -ppmlhdfe- and -ivppml-, I know I should understand, but I'm still unclear how the multi-level fixed effects and dummy endogenous variables can be handled together in this corner solution setting.
Many thanks,
Ziwei
0 Response to ppml with fixed effects and endogenous variables
Post a Comment