2
$\begingroup$

I am trying to implement a quantile IV model, and I must confess that I'm not fully familiar (read: comfortable) with the theory, although I have read the Chernozhukov and Hansen paper. However, the CQIV do code in Stata only provides confidence interval, and no standard errors. Why is this? Is the model such that it is not possible to get standard errors? Surely bootstrap would work, since that is how we get the confidence intervals?

Reference: http://www.sciencedirect.com/science/article/pii/S0304407605000643 Chernozhukov, Victor, and Christian Hansen. "Instrumental quantile regression inference for structural and treatment effect models." Journal of Econometrics 132.2 (2006): 491-525.

$\endgroup$
1
  • $\begingroup$ NB: there are three different approaches to QR with endogeneity, with three different sets of assumptions. None is "better" or "worse," just appropriate for different applications. See Section 1.2.5 in faculty.chicagobooth.edu/christian.hansen/research/… for more (published version: Section 9.2.5 in Chapter 9 in the Handbook of Quantile Regression). I think Stata's cqiv is actually the control function approach, not the "instrumental variables" approach of Chernozhukov and Hansen (2005,2006). $\endgroup$ Aug 29, 2019 at 0:38

1 Answer 1

1
$\begingroup$

For Stata estimation of the Chernozhukov and Hansen (2005) IVQR model, within Stata issue command

net from https://kaplandm.github.io/stata

and follow instructions to install the sivqr command.

The cqiv command is actually based on a different underlying model (a triangular system).

For details on all of this, see the Stata Journal article draft (in 3rd round of review as of 24feb2022) at https://kaplandm.github.io/#sivqr

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.