Unrestricted VAR for granger causality

[iboha]

Dear all,
I have a question regarding VAR.Please lend your hand.
I have two time series that are both I(0). I established a unrestricted VAR in order to test for the causality. However,when I tried to select the optimal lag length using the information criterion in Eviews I faced some problems. All the criterion tend to select the max lags specified by me as the optimal lag length. i.e. if I put 8 for max lag, all choose 8; if I change the max lag to 10, all select 10 as the optimal lag length. Same thing happen while I keep increasing the lags. For example,

VAR Lag Order Selection Criteria
Endogenous variables: Eu Gp
Exogenous variables: C
Date: 04/21/14 Time: 13:09
Sample: 1953 2010
Included observations: 46

Lag LogL LR FPE AIC SC HQ

0 112.3816 NA 2.82e-05 -4.799199 -4.719693 -4.769415
1 198.1069 160.2692 8.09e-07 -8.352476 -8.113958 -8.263126
2 341.5737 255.7451 1.88e-09 -14.41625 -14.01872 -14.26733
3 375.1280 56.89646 5.23e-10 -15.70122 -15.14468 -15.49273
4 475.5361 161.5261 7.95e-12 -19.89288 -19.17732 -19.62482
5 509.3122 51.39832 2.20e-12 -21.18749 -20.31292 -20.85987
6 530.3694 30.21249 1.06e-12 -21.92910 -20.89552 -21.54192
7 545.4312 20.30068 6.67e-13 -22.41005 -21.21746 -21.96330
8 590.1622 56.40005 1.16e-13 -24.18097 -22.82936 -23.67465
9 624.2879 40.06058 3.24e-14 -25.49078 -23.98016 -24.92489
10 637.9245 14.82241 2.22e-14 -25.90976 -24.24013 -25.28431
11 649.7679 11.84338 1.66e-14 -26.25078 -24.42214 -25.56576
12 668.4326 17.04167* 9.37e-15* -26.88837* -24.90072* -26.14379*

* indicates lag order selected by the criterion
LR: sequential modified LR test statistic (each test at 5% level)
FPE: Final prediction error
AIC: Akaike information criterion
SC: Schwarz information criterion
HQ: Hannan-Quinn information criterion

And there appears serial correlation in the residuals. The only solution to me seems that I should add more lags. But my sample size is 58 only(annual). So I really do not know how to solve this problem. This has driven me crazy these days. I cannot eat, sleep well without solving it. Please guide me on this.

Moreover, when we check for serial correlation of the residuals, if we cannot obtain the perfect results, i.e. there will still be some evidence of serial correlation, e.g.,
VAR Residual Serial Correlation LM Tests
Null Hypothesis: no serial correlation at lag order h
Date: 04/21/14 Time: 13:07
Sample: 1953 2010
Included observations: 56

Lags LM-Stat Prob

1 4.738838 0.3152
2 7.899848 0.0953
3 15.88944 0.0032
4 0.954501 0.9166
5 1.473743 0.8313
6 6.413722 0.1703
7 0.594069 0.9637
8 2.178056 0.7030
9 2.349100 0.6718
10 1.160943 0.8845
11 2.149208 0.7083
12 1.466070 0.8326
is it safe to conclude that the VAR is well specified please?

[iboha]

Has anybody encountered this kind of issue before? Or is it that I should not use VAR to test granger causality please? My understanding is that VAR is in line with granger causality. However, if it is not appropriate then is it right that I should just use the Group function of pariwise granger causality please? But then how should we determine the optimal lag length then? After all, in essence, the underlying rationale of granger causality is VAR, am I right? Or should we choose the lag based on our own judgement. Say 3 for annual data?
Please share your knowledge dear all!I need your gudiance!

[iboha]

Any experts or guru please help me out! Thank you so much in advance!

[resbaby]

Can you attach your file(in eview ) and see the problem.