A few thoughts:

 1. Can you not just use a goodness-of-fit test?  Choose a distribution and compare both samples.  Or use a qqplot.  You may want to do this with returns (i.e. changes) instead of the original series, since this is often easier to model.  There are also relative distribution functions (see, for instance, the [reldist package][1]).
 2. You could look at whether the two series are cointegrated (use [the Johansen test][2]).  This is available in the [urca package][3] (and related book).
 3. There many multivariate time series models such as [VAR][4] that could be applied to model the dependencies (see the [vars package][5]).
 4. You could trying using a [copula][6], which is used for dependence modeling, and is available in [the copula package][7].
 


  [1]: http://cran.r-project.org/web/packages/reldist/index.html
  [2]: http://en.wikipedia.org/wiki/Johansen_test
  [3]: http://cran.r-project.org/web/packages/urca/index.html
  [4]: http://en.wikipedia.org/wiki/Vector_autoregression
  [5]: http://cran.r-project.org/web/packages/vars/index.html
  [6]: http://en.wikipedia.org/wiki/Copula_(statistics)
  [7]: http://cran.r-project.org/web/packages/copula/index.html