In the book 'Regular Polytopes' (H.S.M. Coxeter, 1973), Coxeter describes the $n$-orthoplex (which I will refer to from now on as $\beta_{n}$, and is the $n$-dimensional analogue of the octahedron) as the dipyramid with base $\beta_{n-1}$. This description leads to the following equation (Where $N_{k}^{n}$ denotes the number of $k$-simplexes which are elements of $\beta_{n}$)
$$N_{k}^{n}=N_{k}^{n-1} + 2N_{k-1}^{n-1}$$
We have further that $N_{0}^{n} = 2n = 2^{1}\binom{n}{1} \quad \forall n \in \mathbb{N}$.
The author then states the result in the title follows easily by induction, however I am not sure exactly how. Could anyone possibly help?
There is a pretty easy direct argument. The $n$ orthoplex has $n$ pairs of vertices $v_i^{\pm}$ for $i=1,\ldots n,$ and a subset of the vertices represents a face iff it does not contain both a vertex $v^+_i$ and its opposite $v^-_i$. This is easy to see by how the orthoplexes are inductively constructed.
Now a $k$ simplex has $k+1$ vertices, and since it can't contain both vertices in a pair, we must first choose $k+1$ indices which yields $\binom{n}{k+1}$ choices, and then choose on vertex from each pair, which is another $2^{k+1}$ choices.