Comparing two exchange rate models with different dependent variables I want to compare two exchange rate models namely Balassa-Samuelson model and monetary model of exchange rate. Both models are totally different from each other and their dependent variables are also different. I don't know how to compare these two models and which econometric technique I should use for this comparison.
 A: Since the dependent variable is different in each model, then you can only perform "conditional comparison" (the term "conditional" in its general and not probability-related meaning), for which no rigorous methodology exists.
Each model has two purposes : to explain as much as possible of the variation in its own dependent variable present in the available sample, and to forecast as accurately as possible its own dependent variable out-of-sample.
So there is no special evaluation procedure here, just more or less standard statistical tests regarding the adequacy of each specification (separately), like fit, acceptance of linear specification (Ramsey or RESET tests), estimator variance and statistical significance of coefficients, mulicollinearity, reasonable coefficient signs, nicely behaved errors (heteroskedasticity - autocorrelation tests), forecasting error... 
Assume now (although it is not likely that you will get such a clear picture) that one of the two models (say model $A$) performs better on all (or most) of these counts than model $B$. What would that tell you? That model $A$ is "better conditional on its own dependent variable". And where does that leave you?  
You can think of three different cases:  
1) The theoretical arguments supporting the two models are deemed equally reasonable and/or plausible. Then you might as well choose model $A$ to imperfectly model reality since in statistical and econometric terms performs better.  
2) Model $A$ is considered, again on theoretical grounds, a more realistic abstraction of reality. Then you just got lucky. 
3) Model $B$ is the one considered superior as relates to its theoretical foundations. Then you just got unlucky.
The point here is that you cannot use econometric performance as direct proof or indication that one model is superior than the other, because, while in their theory they attempt to model and explain the same real-world phenomenon (exchange rate), in their implementation they end up having different dependent variables.
