Panel data - thesis research

[steven]

Dear members,

In the prospect of obtaining my master degree in business studies I am researching the determinants of the allocation of R&D location in the EU by multinationals. Although we had some basic econometrics at school i'm not very experienced and for sure not in practice.
I'm experiencing some difficulties and I hope you members and eviews users could possibly help me!

I am using panel data from subsidiaries of multinational groups located the EU27 (Cyprus and Malta excluded) between 2005 and 2010.
Total number of observations: +/- 140.000
Dependant variable: procentual change in intangible fixed assets
Independant variable: effective tax rate (ETR), statutory tax rate (STR), earnings before interest and taxes (LnEBIT), labour cost (LnLc), GPD/capita (LnGDPC), Education (Educ), Total assets (LnTA), Employees (LnEmpl).

First of all I am having some problems with outliers in variables like ETR
The effective tax rate varies from 0,00 to 826 while the statutory tax rate is average 0,31 (31%).
When observing the histogram and boxplot of the ETR I conclude there are some outliers that bias the variable strongly!
Now my first question is wether I should trim or winsorize, seen my 140.000 observations.
For this variable I should also only treat the variables (outliers) located in the upper range because an effective tax rate of 0% is a possible case.

Is there somebody that might possibly help me?
Thanks in advance!

[dreea91]

Sorry i'm not able to help you with your problem, i'm a beginner as you are and i have some troubles in finding materials about the winsorizing and trimming procedure.do you have any idea how could i implement one of those procedures to deal with the outliers? thank you