This paper examines the effects of regional R&D on patenting for Sweden within an accessibility framework. We use two measures of patenting: number of patents granted per capita and a composite of quality-adjusted patents which we regard as an innovation indicator, respectively. Three conclusions emerge. First, we find that the specification where innovations per capita is used as a dependent variable performs much better than with granted patents per capita for capturing relationships with regional R&D. In fact, quantile regressions over the distribution of different patenting and innovation levels per capita show that R&D efforts within regions affect innovations per capita positively, except for the regions with the lowest levels of R&D. The effetcs on granted patents per capita are less robust and depend inconsistently on the level of R&D. Secondly, accessibility to inter-regional R&D do not affect innovation significantly in our results, which suggests that effects are locally bounded. This implies that studies of the R&D-innovation relationship are plagued by misspecification, since studies tend to show that R&D-effects diffuse to other regions. This is also the case in our study: the inter-regional effects are an important factor for granted patents. Third, the share of university R&D of all regional R&D has no effect on patenting, which suggests that the two types of R&D are substitutes. In view of these results the redommendation must be to use quality-adjusted patents for regional innovation studies rather than patent grants.