I am considering $k$-fold covers of smooth manifolds (with smooth covering maps).
Let $f:M^m\to N^m$ be a smooth finite covering map.
--
The following implication is not true: $M$ can be embedded into $\mathbb{R}^n$ $\Rightarrow$ $N$ can be embedded into $\mathbb{R}^n$.
The easiest counterexample is $M=\mathbb{S}^{n-1}$ and $N=RP^{n-1}$.
--
But what about the converse of the implication:
$N$ can be embedded into $\mathbb{R}^n$ $\Rightarrow$ $M$ can be embedded into $\mathbb{R}^n$.
Intuitively, this makes sense since if $N$ can be embedded, we should be able to embed $M$ which is obtained by "cutting, unfolding and gluing several copies of $N$" and is therefore "less complicated".
Or is there an obvious counterexample I'm missing?
NB: I'm of course aware of the Whitney Embedding Theorem. The point is, however, to derive something better than $n=2m$ for $M$ based on the fact that I know something about the embeddability of $N$.
The converse is also false. This is because there are planar surfaces that are finitely covered by non-planar surfaces. The first example that comes to mind is the four-holed torus double covering the four-holed sphere (three-holed disk).
In dimension three there are many more examples - for example most finite covers of knot complements do not embed in euclidean three-space.
[Edit] It appears that adding the hypothesis of "closed" doesn't save us -- Section Three of (http://arxiv.org/abs/0810.2346) gives a few examples of closed three-manifolds (of Nil and Solv geometry) that embed in $S^4$ and which have (infinitely many) finite covers that do not embed in $S^4$.