1
$\begingroup$

I have a problem in determining the right statistical test for my research. First I will give some background information. I want to research the impact of IFRS (regulatory change) on audit fees (fees payed to external auditors). This change was mandatory in the year 2005. I will use several pre-adoption years (2000-2004) and several post-adoption years (2005-2008), so I think I can use an OLS regression for this.

However now there is a problem. I also want to test the long term effect of IFRS to see if there is a difference in the years (2008-2012). To make it more specific I will give some examples of my thoughts.

E.g. I can test pre-adoption (2000-2004) to post-adoption (2008-2012)

Pre-adoption (2000-2004) to all post adoption years (2005-2012) (include the years of main hypothesis)

Pre-adoption (2000-2004) to change in audit fees per year, so not necessarily a median change. (this one doesn't lose statistical power over time if I'm correct)

Or use a logit model to see if there is a change yes=1 or no change=0 between the years (2004-2008) and (2009-2012).

I hope you can see my problem from the information above and recommend what is the best possible test to test this effect?

$\endgroup$

1 Answer 1

0
$\begingroup$

Since I perform calculations related to implementation of IFRS, your question is quite interesting to me.

If I understand correctly, you want to test pre vs post adoption, while controlling for natural creep in charges over time.

In this case, I would begin with looking at OLS where the outcome is audit fees, the predictors are binary adoption/no adoption and year, and the model would include interaction terms, and see how it performs. Outcome from such model should make you able to perform some conclusions on isolated effects of the natural creep, and the fact of adoption.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.