Hello everyone,
I am running a logistic panel regression to determine how the enactment of the Sarbanes Oxley Act and the share of independent board members impact the likelihood of a firm to commit fraud in a certain period.
My variables are the following:
lag1SH_INDEP = share of independent board members of a firm in % (between 0 and 1) in the previous period
SOX = dummy variable coded 1 for the years after SOX was enacted and 0 for the years before
FRAUD_P = dummy variable coded 1 if a firm committed fraud in a certain period and 0 if it did not
My goal would be to be able to make statements along the lines of "An X% increase in the share of independent board members results in a Y% decrease of the probability of the company committing fraud in the subsequent period".
I run the following regression and then calculate the margins using the - , eyex() - option which I thought would give me the change of fraud probability in % for a change of independent board members in % according to my understanding of this file: https://www.stata.com/manuals13/rmargins.pdf
However, as I interpret it, that would mean an 83.5% decrease of fraud probability for a 1% increase in independent board member share, which seems unrealistic to me.
Would anyone happen to know which mistakes I made and which command would give me the desired value? It would also be great to know whether - margins, dydx(SOX) - would be the correct command to determine the change in fraud likelihood for the post-SOX period?
I am grateful for any help on this issue. As I am new to this forum, please excuse any errors in formatting or wording my questions, I am glad for any advice on how to improve this as well.
Thank you in advance and best regards,
Mara
I am running a logistic panel regression to determine how the enactment of the Sarbanes Oxley Act and the share of independent board members impact the likelihood of a firm to commit fraud in a certain period.
My variables are the following:
lag1SH_INDEP = share of independent board members of a firm in % (between 0 and 1) in the previous period
SOX = dummy variable coded 1 for the years after SOX was enacted and 0 for the years before
FRAUD_P = dummy variable coded 1 if a firm committed fraud in a certain period and 0 if it did not
My goal would be to be able to make statements along the lines of "An X% increase in the share of independent board members results in a Y% decrease of the probability of the company committing fraud in the subsequent period".
I run the following regression and then calculate the margins using the - , eyex() - option which I thought would give me the change of fraud probability in % for a change of independent board members in % according to my understanding of this file: https://www.stata.com/manuals13/rmargins.pdf
Code:
. quietly xtlogit FRAUD_P i.SOX c.lag1SH_INDEP
. margins, eyex(lag1SH_INDEP)
Average marginal effects Number of obs = 23,390
Model VCE : OIM
Expression : Pr(FRAUD_P=1), predict(pr)
ey/ex w.r.t. : lag1SH_INDEP
Delta-method
ey/ex Std. Err. z P>z [95% Conf. Interval]
lag1SH_INDEP -.8350884 .2050966 -4.07 0.000 -1.23707 -.4331064
Would anyone happen to know which mistakes I made and which command would give me the desired value? It would also be great to know whether - margins, dydx(SOX) - would be the correct command to determine the change in fraud likelihood for the post-SOX period?
I am grateful for any help on this issue. As I am new to this forum, please excuse any errors in formatting or wording my questions, I am glad for any advice on how to improve this as well.
Thank you in advance and best regards,
Mara
