BACKGROUND: Level of education is in many fields of research used as an indicator of social status. METHODS: Using Statistics Denmark's register for education and employment of the population, we examined highest completed education with a birth-cohort perspective focusing on people born between 1930 and 1974. RESULTS: Irregularities in the educational data were found for both men and women born from 1951 to 1957. For the birth cohorts born from 1951 to 1954, a sudden increase in the proportion of persons with basic school education only was seen, and a following decrease in this proportion was seen for the birth cohorts born from 1955 to 1957. For the same birth cohorts, a reverse curve was found for the proportion with vocational training as highest completed education. Using proportion of women with at least one child at the age of 30, our analysis illustrated that spurious patterns may emerge when other social phenomena are analysed by partly misclassified educational groups. CONCLUSIONS: Our findings showed that register data are not always to be taken at face value and that thorough analysis may unravel unexpected irregularities. Although such data errors may be remedied in analyses of population trends by use of extrapolated values, solutions are less obvious in epidemiological research using individual level data.