Think about it like this: Imagine you have a big set [tex]S[/tex], with [tex]n[/tex] elements. You want to find the number of ordered pairs [tex](A, B)[/tex], where [tex]A[/tex] and [tex]B[/tex] are subsets of [tex]S[/tex], [tex]A[/tex] has size [tex]m[/tex], and [tex]A \subseteq B[/tex]. There are two ways to count the number of such ordered pairs. You could choose the elements of [tex]B[/tex] first, by letting [tex]|B| = k[/tex] (where [tex]m \leq k \leq n[/tex]); there are then [tex]\binom{n}{k}[/tex] ways to choose [tex]B[/tex], and [tex]\binom{k}{m}[/tex] ways to pick the set [tex]A[/tex] from the elements of [tex]B[/tex]. Thus, for a given [tex]k[/tex], there are [tex]\binom{n}{k} \binom{k}{m}[/tex] ways to choose the pair [tex](A, B)[/tex], so there are
[tex]
\sum_{k = m}^n \binom{n}{k} \binom{k}{m}[/tex]
such pairs in total.
Alternatively, you could choose the elements of [tex]A[/tex] first. This time, there are [tex]\binom{n}{m}[/tex] ways to pick [tex]A[/tex]. Choosing [tex]B[/tex] then amounts to picking any subset [tex]C[/tex] of [tex]S \setminus A[/tex] which you then "add on" to [tex]A[/tex]: [tex]B = A \cup C[/tex]. There are [tex]2^{|S \setminus A|} = 2^{n - m}[/tex] subsets of [tex]S \setminus A[/tex]. Thus, there are
[tex]
2^{n - m} \binom{n}{m}[/tex]
such ordered pairs. Thus, since both methods count the same thing, you have
[tex]
\sum_{k = m}^n \binom{n}{k} \binom{k}{m} = 2^{n - m} \binom{n}{m} \textrm{.}[/tex]