In 1965, Paul Erdős asked about the largest family $Y$ of $k$-sets in $\{ 1, \ldots, n \}$ such that $Y$ does not contain $s+1$ pairwise disjoint sets. This problem is commonly known as the Erdős Matching Conjecture. We investigate the $q$-analog of this question, that is we want to determine the size of a largest family $Y$ of $k$-spaces in $\mathbb{F}_q^n$ such that $Y$ does not contain $s+1$ pairwise disjoint $k$-spaces. Here we call two subspaces disjoint if they intersect trivially. Our main result is, slightly simplified, that if $16 s \leq \min\{ q^{\frac{n-k}{4}},$ $q^{\frac{n-2k+1}{3}} \}$, then $Y$ is either small or a union of intersecting families. Thus we show the Erd\H{os} Matching Conjecture for this range. The proof uses a method due to Metsch. We also discuss constructions. In particular, we show that for larger $s$, there are large examples which are close in size to a union of intersecting families, but structurally different. As an application, we discuss the close relationship between the Erdős Matching Conjecture for vector spaces and Cameron-Liebler line classes (and their generalization to $k$-spaces), a popular topic in finite geometry for the last 30 years. More specifically, we propose the Erdős Matching Conjecture (for vector spaces) as an interesting variation of the classical research on Cameron-Liebler line classes.