It depends on the context. Generally, the most valid data from crime apparently comes from 'victimization surveys' - as opposed to actual police data, which is

It depends on the context. Generally, the most valid data from crime apparently comes from 'victimization surveys' - as opposed to actual police data, which is politicized. So, from victimization surveys - people can only necessarily be 'suspected' of a crime i.e. 'were you robbed in the last year'? Well, if the person answered 'yes' - then the language may only imply that a specific individual was 'suspected' of a crime. Also, it's pretty difficult to determine ethnicity. So this may not simply be a case of 'police suspects' vs 'police arrests' etc.. It's complicated. Better to have the references.