Professor Pepinsky highlighted an article he wrote, published in 1987, based on an analysis of police recorded crime trends in Sheffield.
In the article he argues that the police 'will record trends they are paid for or threatened to avoid, by incentives which have no demonstrable relation to changes in the behaviour of groups of residents'. Furthermore:
'statistics tell much more about people who compile them than about statistical subjects. "Notifiable offences" are police behaviour... There may be some point in trying to explain how police count crime, but there is no apparent point in asking police to do the counting in the first place... I cannot draw inferences about how much victims suffer from the peculiarities of how police do the recording'.
He concludes that 'counting crime is not a sensible way to do criminology or criminal justice'.
Professor Pepinsky's full article - 'Explaining police-recorded crime trends in Sheffield' - can be read here (paywall).