This is one of my old unsolved questions when I reading Novikov's book on homology theory. I do not know how to prove it because standard triangulation, fundamental diagram, etc does not help and it should be easy to prove.
This question just floated up again, so let me put some references here.
This MO question gives several different proofs that for your surface $S$, $\pi_1(S)$ is free. In fact, Lee Mosher's answer gives a direct proof that $S$ is homotopy equivalent to a graph, and hence to a bouquet of circles. You can also proceed by noting the universal cover of $S$ is contractible, and hence $S$ is homotopy equivlanent to any $K(\pi_1(S),1)$, of which the appropriate bouquet of circles is one.
I would agree with Leonid. The claim should definitely be true for open surfaces which are built from closed ones by removing a top handle (i.e. a disc), (this should follow from a handle decomposition and the classification of closed surfaces). For example (as also stated in wiki if you search for surfaces) you could take a cantor set in the sphere, and take its complement. I don't expect this to have a good homotopy type.