How to deal with high kurtosis in VAR model

External question — Cross Validated Stack Exchange Author: Ranto Antonio Original post: https://stats.stackexchange.com/questions/653969 License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/ Adaptation: HTML converted to plain text; contact email addresses removed. I'm trying to do a VAR model on Stata to find the effect of some variables (Exchange rate, volatility, trade openess, GDP and school enrollment rate) on inflow FDI. While most of them are not stationary at level with the ADF test, they all are after differenciation (at least at lag 0 and 1 : some aren't at lag 2). Since they are at the same order of integration, I decided to do a VAR model but the R² is far too low (19%). When I start doing the diagnosis : there was no autocorrelation of the residuals, the model was stable but the distribution of these residuals was not normal (the jarque bera test had a p-value of 0.0000, the skewness was fine but the kurtosis was 0.000 too) I've tried using log transformation (apart from the FDI since it has negative number on it) but that didn't solve it. So is there any method to lower that kurtosis ? The heteroskedasticity test (Breusch–Pagan/Cook–Weisberg test) was at 0.0599 but I'm not sure if it has anything to do with it