The interview data were derived from the evaluation of the National Supported Work Demonstration, which provides work experience under close supervision, peer-group support, and graduated performance standards for hard-to-employ persons. Arrests were chosen as the indicator of criminal activity because thay can easily be verified through police records. Interview data at project enrollment and 9-month followups for 3 years were gathered from 761 experimental and control subjects. The sample included both ex-offenders and ex-addicts from the Hartford, Conn., Oakland, and San Francisco sites. An analysis of self-reports and official arrest records indicates that, after enrolling in Supported Work, both experimentals and controls underreported their arrests. However, there was no significant difference in misreporting between the two groups. The number of arrests on the police records, particularly arrests for drug offenses, were positively related to underreporting. More accurate reporting was associated with a respondent's receiving a followup interview in prison. Also, the number of arrests erased from police records was positively related to underreporting. Further, the respondent's race was a significant factor in accuracy of reporting. Generally, whites underreported only half as much as black and Spanish-speaking respondents. Appended are arrest record transcription sheets, additional tables, and 12 references.