Econometrics: panel data

Hi,

My problem is the following:
I am currently working with panel data. As you know, in many papers,
the method of estimation consists in calculating three-years means to
eliminate short-term fluctuations. The problem is that it reduces the
temporal dimension. I have used this method but I have also taken
annual data. The problem is that the results differ from one method to
another so, my questions are:
1) According to you, which is the best method (three-years means or
annual data) ?
2) From a theoretical point of view, what could be the consequences of
each method ?

Thank you very much,

Khalti Menana
.