If you have a structural econometric model (e.g. a supply/demand system) where you have explanatory variables that are endogenous) you must use either 2SLS or 3SLS which involve instrumental variables or FIML. Some equations in larger macro models are often estimated using 2SLS.

A VAR is a reduced form and thus problems of endogeneity do not arise. 

Models like 2sls and 3SlS are used for such purposes. 

go through this

http://stats.stackexchange.com/questions/216254/time-series-and-instrumental-variables

An instrumental variable (IV) approach to getting consistent estimates when there is an endogenous regressor (or regressors) requires that there are some variables available that are correlated with the endogenous regressor, but are not correlated with the error term in the model. OLS estimation does not yield unbiased or consistent estimates of the parameters in the spatially autoregressive model, because the error terms are correlated with the spatially weighted dependent variable.

A common approach when confronted with this sort of problem in the linear regression context is to try to use the technique of instrumental variables (IV), otherwise know as two-stage least squares (2SLS). IV does not give unbiased estimates, but does give consistent estimates.

A small comment on Beneberu Teferra - When the explanatory variables are stochastic OLS is also consistent  but not unbiased. 

You will find a time series example, a small 3-level equation urban transport system, in the enclosed paper (available on the ResearchGate site):

Gaudry, M., "A Study of Aggregate Bi-Modal Urban Travel Supply, Demand, and Network Behavior Using Simultaneous Equations with Autoregressive Residuals", Transportation Research B 14, 1-2, 29-58, 1980.

The estimators (2SLQ, 3SLQ and FIM) are compared in a credible time-series situation, each structural equation exhibiting a specific multiple-order autoregressive scheme. Box-Jenkins methods are used to analyse the autoregressive schemes before corrections and after to ensure that final residuals are white noise.

VAR model should be used because it power to holding both endogenous and exogenous variable

Dear respected researcher,

Many thanks for your valuable comments.

I will update soon about the model.

Thanks, 

I have read all the comments above. I disagree with some, but I do not feel like taking that route.  The answer is to use IV. and you may test the validity of your IV. 

I think IV can be used in time series methods as it is used in cross-section studies, but one needs to understand the if their are nuisance parameters and omitted variables. For instance if you are simply using a simple linear model with a time series data ignoring unit roots and long-run relationships, you can adopt IV only if the two most popular assumptions are met. That is, Cov(z,u)=0 and Cov(z,X) not equal to zero.

The simple answer is Yes and you can confirm the relevance of the instrument var by overidentifying restrictions test such as Sargan’s test which is based on T*Rsquared

Md Fuad very late reply...as i am new to research gate..see Introductory Econometrics- A Modern Approach by Woolridge. He discussed this issue very clearly there.

A Vector Autoregression (VAR) is a reduced form. Therefore endogeneity is not relevant. There is no need to use instrumental variables in a VAR.