A panel with year fixed effects, industry fixed effects and clustering by firm should be sufficient given the existing literature. Year by Year regressions are used as an alternative to show that the pooled sample is not misrepresent the results.

How about comparing the results of both panel analysis and year-wise regressions, if you are not in hurry?

If you are planning to combine cross sectional effect with time series, panel is a better choice. I mean if the change in situation from year to year is an important aspect, I'd prefer panel. but if the audit quality and earnings management relation is the only subject you are analyzing, running three yearly regressions is enough I think..

Thank you for your help. In my case, probably the best solution is to run both panel analysis and separate regression models, comparing the results. The problem is to understand what approach a referee prefer...

I am strongly in favor of Panel Data analysis when dealing with time series. But if you have time, try both and compare them.

Hello Marco,

Please share your findings too. I am working on determinants of cash holdings. Will be very happy to understanding the difference between PA and regression models.

Ramana

Hello.

I think both approaches are applicable. However, being only three years in the study, there may be problems with the application of the panel data.

I advise the use of a CrossSection regression that includes the three years studied. I think in this way you can get more robust results.

But if you have time, try both and compare them.

You can share your findings?

Augusta Ferreira

Tank you Ramana, thank you Augusta.

I am bringing to a conclusion my empirical analysis and I am trying both approaches, comparing the results. As soon as possible, I will share findings.

Panel data models are better with robust results as it allow for heterogeneity across groups and introduce individual-specific effects, which multivariate regression will not.

I had been wondering on this question for sometime.