Probit with clustered data

[Henry Morgan]

Hi there! New to Statalist, but I was hoping one of you could help me with a spot of confusion.

I am attempting to apply a probit specification to some cross-sectional data which should display clustering (within villages).

I was using the vce(cl variable) and the score bootstrap in boottest to attain robust standard errors, but I just read that the coefficients of the probit themselves may be biased and inconsistent?

Would clustering produce heteroskedasticity which would bias the probit coefficients?

I would rather not move to an LPM due to the bias issues there, but is there an alternative?

Thank you so much,
Henry

[John Mullahy]

Henry: You might want to take a look at this previous thread

https://www.statalist.org/forums/for...tandard-errors ,

at section 3.1.1. in this paper by Jeff Wooldridge

https://pdfs.semanticscholar.org/ebc...f957d9fbb8.pdf ,

at section VII(c) in this paper by Cameron and Miller

http://jhr.uwpress.org/content/50/2/317 ,

and the references to Wooldridge's textbook in the earlier thread.

[Henry Morgan]

Oh thank you very much John!

I can't believe I missed that thread...

It was a great help.