INTRODUCTION: Retrospective studies show a strong association between self-reported child abuse and subsequent tobacco use. Prospective studies using reports to statutory agencies are less common with limited information on people in their 30s. In addition, there have been no comparisons of the effect of self- and agency-reported abuse on smoking. We therefore assessed the effect on the prevalence and persistence of smoking at the 30-year-old follow-up of prospective agency notifications of child abuse compared to retrospective self-reports of maltreatment in the same birth cohort. METHODS: There were 2443 young adults with data on smoking and child abuse at 30-year-old follow-up. Information on self-reported abuse was collected using the Child Trauma Questionnaire (CTQ) and linked to notifications of child maltreatment to statutory agencies. RESULTS: The prevalence of self- and agency-reported maltreatment was 600 (24.7%) and 142 (5.8%), respectively. At follow-up, 565 participants smoked (23.1%) but only 91 (3.8%) smoked 20 or more cigarettes a day. Of the 206 participants who smoked at 14 years, 101 were still smoking at follow-up. On adjusted analyses, both self- and agency-reported maltreatment showed a significant association with the prevalence and persistence of smoking from 14 years old. However, associations were weaker for some of the agency-notified child maltreatment subtypes possibly because of lower numbers. CONCLUSIONS: Child maltreatment is associated with both an increased prevalence and persistence of smoking at 30-year-old follow-up irrespective of reporting source. This is despite self- and agency-reported maltreatment possibly representing different populations. Smoking cessation programs should therefore target both groups. IMPLICATIONS: Retrospective studies show an association between self-reported child abuse and subsequent tobacco use. Prospective studies of reports to statutory agencies are less common with limited information on people in their 30s. In addition, there have been no comparisons of self- and agency-reported abuse on smoking outcomes even though they may represent different populations. We therefore compared the effect of both on smoking outcomes at 30-year-old follow-up of 2443 adults from the same birth cohort. On adjusted analyses, both self- and agency-reported maltreatment showed significant associations with the prevalence and persistence of smoking. Smoking prevention and cessation programs should therefore target both groups.