Criminal careers in self-reports compared with official records.

One of the most basic questions in criminology is how to measure the key outcome variable of offending. The two most common measurement methods are to use official records of arrests/convictions or self-reports of offending. These two methods have different advantages and problems. For example, official records include only the ‘tip of the iceberg’ of offending, but the offences are recorded quite accurately and completely for very large samples. Self-reports, on the other hand, reveal more of the ‘dark figure’ of offending but they have problems of exaggeration and concealment and also problems of missing data in interviewed samples compared with target samples. The most delinquent people are often the most difficult to interview (Farrington et al., 1990). A key question concerns the agreement between official records and self-reports. Although self-reports reveal more offenders and offences, to what extent do the worst offenders according to self-reports coincide with the worst offenders according to official records? And to what extent are risk factors similar for self-reported and official offending? If a risk factor predicts official but not self-reported offending, is it predicting the likelihood of an offender being convicted? (see e.g. Farrington, 1979; West and Farrington, 1973). If a risk factor predicts self-reported but not official offending, is it predicting the likelihood that an offender will admit an offence? There have been many studies of the validity of self-reported offending in relation to official offending (see Jolliffe and Farrington, 2014). In most cases, officially arrested or convicted offenders admit their arrests or convictions and admit the types of offences that led to these official records; this is termed ‘concurrent validity’. More impressively, self-reports of offending by currently unrecorded persons predict their future arrests or convictions for the same offences; this is termed ‘predictive validity’ (Farrington, 1989). According to commonly accepted psychometric criteria, self-reports of offending are reliable and valid, at least for young males.