VAR Forecast

[brennan6738]

Can someone possibly take a look at my VAR (out of sample) forecasts and tell me how I might improve upon them? I am trying to model a BBB index yield using four different variables (I have entered all variables as endogenous). The data has 1,750 observations (the data is daily business day data). I forecast using models of order 5, 20 and 120. None of the forecasts seem appropriate. The models of order 5 and 20 seem to capture the actual downtrend at the end of the actual data (but only after forecasting a spike that never occured). None of the forecasts is even within a percentage point of the actual yield at the samples end. I thought the model of order 120 would work best (with 1,750 observations I thought it would not affect the degrees of freedom). Do I need to try a different setting when I make the model? When I solve the model I use the Deterministic/Static option (and leave all other settings unchanged). Do I need to smooth my forecast? If so, how do I do that? Any suggestions anyone has would be greatly appreciated.