Aim The main aims of this article are to estimate the number of offences that are committed for every one that leads to conviction, and to estimate the probability of an offender being convicted. Method In the Pittsburgh Youth Study, 506 boys were followed up from age 13 to age 24 years, in interviews and criminal records. Self‐reports and convictions for serious theft, moderate theft, serious violence and moderate violence were compared. Results On average, 22 offences were self‐reported for every conviction. This scaling‐up factor increased with age and was the highest for moderate theft and the lowest for serious theft. The probability of a self‐reported offender being convicted was 54%. This percentage increased with the frequency and seriousness of offending and was always higher for African American boys than for Caucasian boys. These race differences probably reflected differences in exposure to risk factors. Conclusions More research is needed on scaling‐up factors, on frequent and serious offenders who are not convicted, on self‐reported non‐offenders who are convicted and on why African American boys are more likely than Caucasian boys to be convicted. Copyright © 2014 John Wiley & Sons, Ltd.