how to deal with seasonality in the panel data

[bruce fan]

Hi folks,

This may be a stupid question, but still hope you guys give me some answers to clarify it. Thanks a lot!

I am doing a panel data regression, and want to put the i.year or i.quarter in the regression to deal with the issue of seasonality, I know usually people say the i.quarter is enough to get rid of the seasonality in the regression. But I want to find out if I put both i.year and i.quarter in the regression what problems there will be? And if I only put the i.year in the regression, then does this mean the seasonality cannot be handled? Thank you again!!!

[Clyde Schechter]

If by seasonality you mean periodic fluctuations within years, then that is handled by monthly or quarterly indicators (or sometimes even weekly). Adding only i.year indicators to the regression would only capture "seasonal" variation that occurs over much longer time periods such as decades and is negligible within a year. Using both year and quarter indicators would both capture seasonal variation within years (at least at the coarseness of quarterly estimation) and would also pick up year-specific shocks or, possibly, longer term time trends at scales longer than a year.