This paper studies the performance of admission control in cognitive radio networks (CRNs). We propose a CRN architecture featuring cooperation among several CRNs in the same geographical area. Joint Admission Control (JAC) enables secondary users (SUs) to have access to the combined spectrum pool of the cooperating CRNs. Three joint admission control schemes are investigated and quantitatively analyzed using continuous-time Markov chain analysis. Analytical results reveal new insights that the channel-aware admission control scheme achieves the lowest blocking probability at the expense of communication overhead for obtaining channel usage information in each CRN while the weighted selection scheme obtains the lowest forced termination probability. Moreover, we quantify the gain of cooperation through performance comparison between Joint Admission Control and separate admission control where SUs are restricted to using only one specific CRN. We demonstrate that JAC can achieve significant performance improvement.