Although the benefits of clustering for innovation have received much attention in the theoretical as well as empirical literature, analyses at the regional level often disregard the characteristics of local firms. We tackle both at the same time: agglomeration externalities (Marshall, Porter, Jacobs) from census microdata, and firm data from the Community Innovation Survey. Importantly, we allow for sectoral heterogeneity of agglomeration forces. We find that the firm characteristics, including those that proxy for 'absorptive capacity', have a much stronger relationship with the propensity to innovate than regular agglomeration externalities. The latter are only statistically significant for a few specific sectors, and even then only for some types of innovation. Sorting of innovation-prone firms into specific locations might therefore be much more important to explain spatial patterns of innovation than agglomeration externalities.