2sls and 3sls, when endogenity because of omitted variable and reverse causality

[Aswer Mohamed Ambrella]

I am writing this blog and would like to get somebody's help me with 2sls and 3sls in stata.
Here is my question Suppose we investigate the link between Corporate social responsibility and corporate financial performance of several firms. We assume that OLS estimates are biased. So, we decide to go for 2SLS as it can control endogenity problem. Suppose we have panal data (balance dataset from 2000 to 2005).

Suppose we write our OLS as

FP (financial performance) = CSR index + Size + Risk + error (1)

As you explained in your papers, this OLS model still suffer from endogeneity problem due to (1) unobservable omitted variables and (2) reverse causality.

So we use 2sls to address this endogenity issue. We think CSR in above equation is endogenous and we do not have propoer instrument to it. But since we have panel data, we take lag value of CSR as our instrument. So we write our 2sls as

sr index = FP + CSR (t-1) + error (2)

FP = CSR index + CSR (t-1) + Size + Risk + error (3)

My question is, how we write 3sls equation in this case and how to write stata codes for 2sls and 3sls for this example. Could somebody help me?

[Phil Bromiley]

You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. By the way, this is a listserve not a blog. It is also Stata not stata.

You refer to "you explained in your papers" - to whom do you think you are communicating? There are a great many folks on this listserve.

If there is serial correlation, a one period lag on CSR is not exogenous. You can't do 3SLS on one structural equation. Read an econometrics text, and look at ivreg and reg3 documentation that comes with Stata.