question about PSM

[Hu Jianhao]

Hi, everyone!

I am investigating the relationship between the independent variable X and ESG performance. My data is such that not all firms disclose ESG reports, but there is a third-party rating agency that give ESG performance scores for all firms, and I use this ESG score as my dependent variable.

However, the reviewer believes that there is a potential selection bias here in that there may be a significant difference in ESG scores between companies that disclose ESG reports and those that do not, so they recommend that I use either the PSM or the Heckman model to address this bias.

Is it possible to use PSM in this case? Match firms that disclose ESG reports with firms that do not disclose ESG reports and then regress X and ESG performance?

(I don't think PSM can solve this problem because I think whether or not one chooses to disclose an ESG report is not the independent variable that I care about. Is my understanding wrong?

[Hu Jianhao]

help...

[Joseph Coveney]

. . .the reviewer believes that there . . . may be a significant difference in ESG scores between companies that disclose ESG reports and those that do not,

Well, is there? You have the "ESG performance scores for all firms".

dotplot <ESG performance score>, over(<firm discloses or not>) median bar center
// if too many, then
tabstat <ESG performance score>, by(<discloses or not>) statistics(q n)

. . .whether or not one chooses to disclose an ESG report is not the independent variable that I care about.

Then take a stand. If you want to be diplomatic, then you can try something like

regress <ESG performance score> i.<discloses or not>##c.X

or whatever method you're using to investigate "the relationship between the independent variable X and ESG performance".

[Hu Jianhao]

Well, is there? You have the "ESG performance scores for all firms".

dotplot <ESG performance score>, over(<firm discloses or not>) median bar center
// if too many, then
tabstat <ESG performance score>, by(<discloses or not>) statistics(q n)

Then take a stand. If you want to be diplomatic, then you can try something like

regress <ESG performance score> i.<discloses or not>##c.X

or whatever method you're using to investigate "the relationship between the independent variable X and ESG performance".

Thank you so much ! I will show the dotplot and tell him I don't think the PSM is appropriate, thanks!