VECM and forecasting

[pkpio24]

Hi,

I have arrived at an equation via the cointegration test process and eliminating insignificant variables:

D(LN_NBP) = C(1)*( LN_NBP(-1) - 0.860929243883*LN_BRENT(-1) +
0.266786377636 ) + C(4)*D(LN_BRENT(-1)) + C(6)


When trying to forecast EViews only forecasts one step forward (I have data jan2000-dec2014 and it only forecasts Jan2015).
Is it something wrong with the equation obtained through the VECM or am I choosing a wrong type of forecast.

LN_NBP is gas price
LN_Brent is oil, and I am trying to forecast the cointegrated future prices.

Output from equation:

-----------
Dependent Variable: D(LN_NBP)
Method: Least Squares
Date: 04/06/15 Time: 16:18
Sample (adjusted): 2000M03 2014M12
Included observations: 178 after adjustments
D(LN_NBP) = C(1)*( LN_NBP(-1) - 0.860929243883*LN_BRENT(-1) +

0.266786377636 ) + C(4)*D(LN_BRENT(-1)) + C(6)

Coefficient Std. Error t-Statistic Prob.

C(1) -0.355109 0.053506 -6.636756 0.0000
C(4) -0.500458 0.170152 -2.941237 0.0037
C(6) 0.009789 0.016654 0.587764 0.5574

R-squared 0.210504 Mean dependent var 0.007435
Adjusted R-squared 0.201481 S.D. dependent var 0.248215
S.E. of regression 0.221805 Akaike info criterion -0.157328
Sum squared resid 8.609538 Schwarz criterion -0.103702
Log likelihood 17.00219 Hannan-Quinn criter. -0.135581
F-statistic 23.33024 Durbin-Watson stat 2.238876
Prob(F-statistic) 0.000000



----------

Please help and thank you very much!

[EViews Gareth]

Do you have data for LN_Brent for the future?

[pkpio24]

No I have not. How should I approach this?

[EViews Gareth]

You've stated that LN_NBP depends on LN_BRENT. If you want to forecast LN_NBP, you'll need data for LN_BRENT.

[pkpio24]

Thank you for clarification, I am quite new to this.
I have data for both until 2014, how would I forecast the future LN_Brent and LN_NBP if they are cointegrated?

[EViews Gareth]

Estimate a VEC, rather than a single equation cointegration. Details of estimating and forecasting VECs can be found in the User Guide.

[pkpio24]

Thanks a lot! It worked very well.
As I have a sample from 2000-2014 how far would you recommend to forecast? Are there any "rules", "rule-of-thumb" or guidelines on how many periods/years one should estimate?