Study 1
Heckman and Karapakula (2019a) reported findings from a 50-year follow-up evaluation on the sample of children from the HighScope Perry Preschool study, which was originally conducted from 1962 to 1965. Researchers collected data annually when the children were between the ages of 3 and 15. Follow-up data were collected at ages 19, 27, 40, and 55. This report includes the study findings collected at the most recent follow up, when the participants reached age 55.
Children were initially identified after a fall survey, conducted each year from 1962 to 1965. Project staff created pairs of children matched on initial Stanford–Binet IQ tests and randomly assigned each pair to one of two undesignated groups. They then exchanged several similarly ranked pair members so the two groups would be matched on socioeconomic status, intellectual performance, and percentage of boys and girls. Once the two groups were identified, they were randomly assigned, through the flip of a coin, to either the program condition or no-program condition.
The total number of study participants was 123. Across the sample, mothers had a mean age of 31 years, an average of 9.2 years of education, and typically worked at unskilled jobs. Fathers had a mean age of 35 years, an average of 8.3 years of education, and typically worked at unskilled jobs. Forty-eight percent of fathers lived at home. Perry families had an average of 4.5 children (Weikart, Bond, and McNeil 1978; Heckman et al. 2010). Of these children, 58 were assigned to the program group, and 65 were assigned to a control group that did not participate in a preschool program. All the children were African American, of low socioeconomic status, and an age of 3 or 4 years. Their IQ scores were low (between 70 and 85, the range for borderline mental impairment), but they had no organic deficiencies (i.e., biologically based mental impairment). They were at high risk of failing school. Approximately 58 percent of the sample was male. There were no differences between the groups with regard to father absence, parent education level (ninth grade, on average), family size, household density (i.e., 1.4 persons per room), or birth order. At the baseline, the only statistically significant difference between groups was in the greater number of employed mothers in the participant group, compared with the comparison group (Schweinhart 2013).
Attrition in the sample was low at each follow-up period. At the follow-up period for this study, 83 percent of the original sample (102 out of 123) completed a new survey. The authors examined outcome measures analyzed in past studies, such as self-reported employment outcomes and measures of criminal activity, obtained from administrative records. Additionally, new measures on cognition, personality, and biomarkers of health were collected through epidemiological exams. Worst-case approximate randomization tests were used to analyze the outcomes. In addition, the authors conducted subgroup analyses on gender.
Study 2
Heckman and Karapakula (2019b) conducted a study on the intergenerational treatment impact of the Perry Preschool Project by looking at the children of the original program participants. The authors collected longitudinal data on the first generation of Perry participants; the midlife surveys administered in 2014, 2015, and 2016 included questions about their children (second generation). Ages of the children across all the years in which the interviews took place ranged from 1 through 43, with a mean of 28 years. Because the survey spanned multiple years, information about children born in the same year was not necessarily collected in the same interview year. In contrast, at the previous follow-up period in 2000, about 63 percent and 80 percent of the children were younger than ages 18 and 21, respectively.
Outcome measures for the children of the original Perry Preschool participants included school suspensions, education, arrests, addiction, and employment. Suspension, arrests, and addiction were analyzed individually and collectively. For this CrimeSolutions review, the pooled effect size for the three outcomes was scored. The authors used an augmented inverse propensity weighting (AIPW) estimator to calculate the intergenerational treatment effects on the outcomes. They also conducted subgroup analyses on gender.