Background Reliable epidemiological data on sexual homicide are sparse, especially on trends in its incidence over time and age at arrest. Aims Our main aims were to study age at arrest and incidence trends for sexual homicide in the USA over about three decades (1976–2007). Methods We conducted longitudinal analyses of data from the largest USA homicide database available for the years 1976–2007. Results The mean age at arrest for a sexual homicide was 26.3 years (range 7–76; modal 21 years). Three quarters of these offenders were young adults aged 18–35. Age at probable first arrest for a sexual homicide rose significantly from a mean of 25 to a mean of 29 years over the study period. The last decade of the three studied accounted for just one quarter of the homicides as charged in the whole period, but the proportion of sexual homicides specifically fell with each decade, so that the first period accounted for 56% of those in the whole period, the second for 33% and the final decade for just 11%. This was reflected in a reduction in the proportion of all homicides that were sexual, from 1.4% in the first decade to 0.8% in the second and 0.4% in the third, declining by a factor of five for adults and seven for juveniles. Conclusions Use of official national criminal statistics has limitations in studying the epidemiology of any particular behaviour. Nonetheless, our findings of falling sexual homicide rates and of changes in at least one important demographic of these killers indicate a need for a considered reappraisal of such crimes. Establishment of accurate epidemiology and a study of associated factors may assist in the improvement of investigative and preventive strategies. Copyright © 2014 John Wiley & Sons, Ltd.