autocorrelation, thesis help please

[matteo.bianchi11]

Hi everybody,
I need help for my thesis please.
I have a three variables var, with inflation, unemployment rate and interest rate. I have estimeted a 4 lags VAR on the period 1960 - 1979 with quarterly data.
If I don't add 2 dummy variables, the residual analysis shows that there is no autocorrelation, but I have not normal distributed residuals.
So I added 2 dummy variables and now I have normal distributed residuals but autocorrelation at lag 3, as in output below:

VAR Residual Serial Correlation LM Tests
Null Hypothesis: no serial correlation at lag order h
Included observations: 78

Lags LM-Stat Prob
1 12.81786 0.1710
2 9.783210 0.3683
3 29.78867 0.0005
4 14.65985 0.1007

How can I remove autocorrelation at this lag?
I heard about treat it as an AR(p) process, but sincerely I don' t know how to do it and what it means deeply.

Is someone available to help my please?
thanks in advance

[nishantvats12]

After adding the dummy variable re-do the analysis of AIC/SIC to determine the optimum lag length before going for a VAR system.

PS: Non normality of residuals is not a serious condition and can be neglected.