The balloon analogy, while it's a lot less wrong than even simpler models, is kind of unfortunate since it has two defects:
- you can immediately visualise the balloon as a two-dimensional object embedded in three-dimensional space and thus you get the question you are asking;
- the balloon is finite, and its surface is curved (intrinsically curved I mean).
Here's a model which is less easy to visualise but gets away from both of these problems.
Consider an infinite two-dimensional plane (pretty much $\mathbb{R}^2$), with a regular rectangular grid inscribed upon it (a coordinate system, if you will). Over time this plane is expanding in the same way that the balloon is expanding, so the spacing of the grid increases over time, or equivalently any two points you were to mark on this plane would get further apart over time.
Now this model is harder to visualise than the balloon because of the whole infinite thing: how do you even think about an infinite object expanding? And it has properties which are hard to understand: for instance two points on the plane have recession velocities proportional to their separation, and since there is no upper bound to the possible separations between points (the plane is infinite) there is also no upper bound to the recession velocities of two points. In particular, this means (if the speed of light is finite) that the plane has a horizon: you can't see points on it further away than a certain distance because they are receding from you faster than $c$ (the universe does have such a horizon in fact). And finally, it's also not completely right as a model because it's possible to imagine such a plane as existing into the arbitrary past, which is not the case for the universe.
But it gets away from two of the problems of the balloon analogy: there clearly is no notion of a 'centre' of this thing about which it is expanding, and neither is the plane intrinsically curved.