Since its beginnings, comparative criminology has suffered from the lack of valid international data on crime and criminal justice. Over several decades, efforts were made to improve international comparisons by new data collections. In this paper, three such data sources are compared. The results show that International Crime (Victim) Survey (ICVS) data are highly correlated with police data published in the European Sourcebook of Crime and Criminal Justice Statistics, at least for offences whose survey and police definitions closely match. On the other hand, Interpol statistics are only weakly correlated with ICVS and Sourcebook data. It is argued that the careful screening process within the Sourcebook project increases considerably the comparative validity of police data by the elimination of often trivial but devastating errors. Despite such improvements, survey data may remain more valid for cross-sectional analyses on personal crime and other offences where recording by the police is uncertain and often inconsistent. In any case, unchecked police data (as those published in Interpol Statistics) should not be used for comparative purposes.