In the past 25 years, the cancer surveillance community has revised several times the coding systems used to record cancer stage at diagnosis. Most recently, the Collaborative Staging (CS) System replaced the directly coded Surveillance, Epidemiology and End Results (SEER) Summary Staging 2000 (SS2000) system for cancer cases diagnosed in 2004 and after in the United States. SS2000 can be derived from CS, producing the so-called Derived SS2000. But CS and SS2000 have not been used to code the same cancer cases simultaneously. Checking the agreement of these two systems has proved challenging. When comparing the stage of cancer cases diagnosed before and after January 1, 2004, observed differences in stage distributions could be attributed to true changes in stage trends, to decreases in the percentage of cases with unknown stage due to improved staging schema in the CS, to differences in the staging instructions between the two staging systems, or to other factors. This paper proposes a method to check the agreement of the two staging systems by comparing the cancer stage distributions, when cases in 2001-2003 are staged using SS2000 and cases in 2004 by CS. We build models to separate the impact of non-coding factors, such as natural trends, from those related to changes in coding instructions. If the non-coding factors satisfactorily explain the differences in stage distributions from pre-CS to CS diagnosis years, then we may conclude that the two systems have no significant coding differences. Otherwise, we provide directions for determining the nature of the discrepancies. Cancer data from the North American Association of Central Cancer Registries are presented to illustrate the method used in checking the agreement between the CS and SS2000 coding systems.