To study how police report crime data in Calgary and Edmonton, administrative procedures and offense-classifying practices (coding) were examined. To test changes in coding over time, police personnel were ordered to recode previously coded cases, without knowing the original code. The samples of cases were randomly drawn and amounted to 152 for Edmonton, 143 for Calgary. Results showed an 8 to 14 percent change in offense classification in Edmonton and 1 to 6 percent in Calgary, mostly in sex offense cases. Moreover, in a test of differences in coding practices between the two police departments, an exchange of previously coded cases for recoding revealed a 2 to 8 percent change of Edmonton cases in Calgary and 11 to 15 percent change in Edmonton of Calgary cases. Again, the majority of recodings involved sex offenses. Personnel from both police departments interviewed about administrative directives for coding indicate difficulty in interpreting Canadian government crime recordkeeping rules and offense classes. Moreover, administrative procedures in crime reporting, in which Calgary police may defer reporting less serious offenses but Edmonton officers may not, results in differences between the two cities in reported assaults. However, tests of simple coding error in Edmonton and Calgary reveal that coding in both cities is reliable. It is concluded that government sex offense classes need clarification, and that the Edmonton-Calgary example shows that crime data cannot be compared for different cities. Footnotes and tables are included.