15
$\begingroup$

I am trying to use instrumental variables analysis to infer causality with observational data.

I have come across a two-stage least squares (2SLS) regression which is likely to address the endogeneity issue in my research. However, I would like to first stage to be OLS and second stage to be probit within the 2SLS. Based on my reading and search, I have seen researchers use either 2SLS or first stage probit and second stage OLS, but not the other way round which is what I am trying to achieve.

I am currently using Stata and ivreg command in Stata is for a straight 2SLS.

$\endgroup$
12
$\begingroup$

Your case is less problematic than the other way round. The expectations and linear projections operators go through a linear first stage (e.g. OLS) but not not through non-linear ones like probit or logit. Therefore it's not a problem if you first regress your continous endogenous variable $X$ on your instrument(s) $Z$, $$X_i = a + Z'_i\pi + \eta_i$$ and then use the fitted values in a probit second stage to estimate $$\text{Pr}(Y_i=1|\widehat{X}_i) = \text{Pr}(\beta\widehat{X}_i + \epsilon_i > 0)$$

The standard errors won't be right because $\widehat{X}_i$ is not a random variable but an estimated quantity. You can correct this by bootstrapping both first and second stage together. In Stata this would be something like

// use a toy data set as example
webuse nlswork

// set up the program including 1st and 2nd stage
program my2sls
    reg grade age race tenure
    predict grade_hat, xb

    probit union grade_hat age race
    drop grade_hat
end

// obtain bootstrapped standard errors
bootstrap, reps(100): my2sls

In this example we want to estimate the effect of years of education on the probability of being in a labor union. Given that years of education are likely to be endogenous, we instrument it with years of tenure in the first stage. Of course, this doesn't make any sense from the point of interpretation but it illustrates the code.

Just make sure that you use the same exogenous control variables in both first and second stage. In the above example those are age, race whereas the (non-sensical) instrument tenure is only there in the first stage.

$\endgroup$
  • $\begingroup$ Thanks so much, this resolved the issue I was facing. Thanks again. $\endgroup$ – Veronica Dec 26 '14 at 16:22
  • 1
    $\begingroup$ Actually research suggests to use a control function approach for nonlinear models like a logit, which would be to use the residuals from the first stage along with the endogenous variable, rather than the predicted values. Though there do seem to be advancements often, see: stat.wharton.upenn.edu/~zijguo/… $\endgroup$ – robin.datadrivers Feb 23 '17 at 3:46
  • 1
    $\begingroup$ Sorry I didn't provide a citation for comparing the two approaches for linear and nonlinear models. ncbi.nlm.nih.gov/pmc/articles/PMC2494557 $\endgroup$ – robin.datadrivers Feb 23 '17 at 5:36
  • $\begingroup$ -1 I had to downvote this answer, as it seems not generally possible to carry over the idea of the 2SLS estimator to nonlinear models in the first and/or second stage. It may be true for the case of 1st stage LS and 2nd stage probit (@Andy do you have a reference to support this?), but at least a caveat is in order as I have seen many people doing the 2SLS idea in all sorts of cases of nonlinear models in first and second stage and that is problematic practice. $\endgroup$ – Momo Sep 18 '18 at 11:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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