Hello,
I am using balanced panel data for the model where the dependent variable is a fraction. Both endogenous and instrumental-variable are binary (0/1). I have read the paper "Panel data methods for fractional response variables with an application to test pass rates," in which the authors considered the endogenous variable as continuous and modeled accordingly. However, in my case, it is not the same. Therefore, I would appreciate advice on expressing the conditional mean model and reduced form model presented in the above-mentioned paper and how I should estimate the residuals from first-stage regression.
Thank you very much in advance.
Kind regards,
Faruque.
There are "fancier" things you could do, but if I were looking for a practical, straightforward solution and facing your data limitations (fractional dependent, binary endogenous regressor, binary instrument), I would start by treating the binary endogenous variable the same as any other in the first stage. Other methods may be able to eek out some efficiency, but (assuming you have a "good" instrument), that should resolve your bias problem. If you have enough data, the efficiency losses are less important.
Dr. Burke, Thank you so much for your valuable suggestion. The dataset contains seven years panel data that includes 2835 observations and the instrument is good. Thanks again.
It is hard to give advice without knowing what you want to analyse and what variables you use. You should show your specification. It seems to be a rather strange model, where a number (a fraction of two integers or real numbers?) is dependent on two binary ones.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science
- Materials
- Inorganic Materials
- Metals
- Alloys
- Steel