Excess Kurtosis

[WNP]

Hey, I'm running a regression on Eviews 6, regressing Inflation, investment and GDP lagged by 1 and 2 periods, on GDP growth per capita, using panel data with fixed effects for time and cross section.

However the residuals have a non normal distribution because they have a kurtosis of 6 (excess of 3), so the Jarque-Bera p-value is 0.0000. My sample size is about 1250.
What I'm trying to find out is:

1) Is there any way to fix this?
2)If not, what does this imply for the p-values of the coefficients obtained from the regression, and inferences in general as it breaks the OLS assumption of normal distribution of residuals.
3) Are Unit root tests on my data reliable given they don't have a normal distribution (mainly inflation)

Any help appreciated
Will