Question about the effect of adding control variables on estimate

[Ke Lyu]

Dear all,

I do a research using weak instrumental variables, the main problem is that we are not sure which result should be chosen . We found the instruments are weak, so we decided to use CLR test, K test and AR test to get the weak-identification-robust inferences. The null hypothesis of these tests is that if estimate of beita is equal to zero. Our estimate of ß is around 0.06.
BUT we found that If we add some control variables, we will accept the null hypothesis, however, if we drop these two control variables, we will reject the null hypothesis.
We found that the two control variables are not related with the explanatory variables, so if we can say that it is because these control variables dilute the strength of estimate of ß which is very small ? Or there are other explanations we can use for this case.

Thank you.

Result of not adding two unrelated control variables

Result of adding two unrelated control variables

[Phil Bromiley]

You didn't get a quick answer. 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. While you provide some results, you don't tell us precisely what procedure produced them, and we can't see the other results (on ivreg or whatever) or replicate the issue.