I am studying earnings management and audit quality in SMEs, analyzing financial statements of a sample of 386 firms for three years. Do you think is more appropriate a panel analysis or to run three regression models (one for each year)?
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.
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...
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.
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.