Previous research studies the age profile of patent citations to learn about knowledge flows over time. However, identification is problematic because of the collinearity between application year, citation year, and patent age. We show empirically that a patent's citation clock does not start until it issues, and propose a highly flexible identification strategy that uses the lag between application and grant as a source of exogenous variation. We examine the potential bias if our assumptions are incorrect, and discuss extensions into other research areas. Finally, we use our method to re-examine prior results on citation age profiles of patents from different technological fields and application year cohorts.