EViews.com

Dear All,
I'm running some time series data, and planning to do classic OLS. but the problem is my variable time series data are not normally distributed. can you advice me what to do with it?
these are some histogram detail about my variables
Series: AVG_VAR
Sample 1 156
Observations 156

Mean 0.000662
Median 0.000412
Maximum 0.006918
Minimum 7.97e-05
Std. Dev. 0.000844
Skewness 4.201737
Kurtosis 25.60200

Jarque-Bera 3779.547
Probability 0.000000
--------------------
Series: R_IHSG
Sample 1 156
Observations 156

Mean 0.001410
Median 0.001933
Maximum 0.070136
Minimum -0.109540
Std. Dev. 0.017897
Skewness -1.275486
Kurtosis 12.79372

Jarque-Bera 665.7592
Probability 0.000000

Thanks in advance,

Ilham

Run OLS.
Normality of the variables is completely irrelevant.

i need all variable in OLS to be normal distributed right?

No.

Normality of the dependent matters in finite samples. The explanatory variables don't matter as long as they have finite second moments.

i thought the classic OLS assumption all variable involve is normally distributed?

Nope, with the quintessential example being dummy variables, which are far from normally distributed, yet still perfectly valid.

All that matters is that the error terms are normal. And that doesn't matter much. (As a minor quibble, the dependent variable need not be unconditionally normal. )

im running 2 variable right now
dependent is excess return and independent is average correlation
these 2 variable is stationer but and i regress it using OLS (newey-west). the result indicating the residual is not normal, so the estimation output consider to be biased dont it? how can i solve this problem?

No, this does not cause bias.