Hi all geniuses,
Please imagine such a situation:
The dependent variable of interest is the creditworthiness of borrowers, and we have designed five measures of borrowers' creditworthiness. In other words, we have five versions of dependent variables, Y1 to Y5.
The explanatory of interest is local banking competition. Through the main OLS regressions, we have found all measures of creditworthiness are effective and show consistency.
Now, to test whether our explanatory variable is suffered from endogeneity, we design an IV approach.
We have designed two IV candidates, let's say IV1 and IV2.
However, we find that for Y1, Y2, Y3, we should choose IV1 and log(IV2) to pass various tests on the relevancy and exogeneity of IVs; for Y4 and Y5, we should choose log(IV1) and IV 2 to pass such tests.
My question is, should we insist on using the same settings of IV1 and IV2? After all, IV approach aims to identify and extract the exogenous part of explanatory variables. So, does the setting consistency really matter?
Thank you all for your generous assistance!
Please imagine such a situation:
The dependent variable of interest is the creditworthiness of borrowers, and we have designed five measures of borrowers' creditworthiness. In other words, we have five versions of dependent variables, Y1 to Y5.
The explanatory of interest is local banking competition. Through the main OLS regressions, we have found all measures of creditworthiness are effective and show consistency.
Now, to test whether our explanatory variable is suffered from endogeneity, we design an IV approach.
We have designed two IV candidates, let's say IV1 and IV2.
However, we find that for Y1, Y2, Y3, we should choose IV1 and log(IV2) to pass various tests on the relevancy and exogeneity of IVs; for Y4 and Y5, we should choose log(IV1) and IV 2 to pass such tests.
My question is, should we insist on using the same settings of IV1 and IV2? After all, IV approach aims to identify and extract the exogenous part of explanatory variables. So, does the setting consistency really matter?
Thank you all for your generous assistance!
Comment