The effect of early noncognitive skills on social outcomes in adolescence

Abstract

This paper investigates the impact of early noncognitive skills on social outcomes in adolescence. The child’s attention span, approach, prevailing mood and distractibility in early childhood may be crucial predictors for school achievements, health risk behavior, delinquency and autonomy as adolescent. We investigate this issue using a longitudinal epidemiological cohort study of 384 children at risk from the Rhine–Neckar region in Germany. Our results indicate that noncognitive skills in early childhood are important predictors of educational success, tobacco and alcohol use, delinquency and autonomy in adolescence. In particular, the attention span has emerged as a dominant factor among noncognitive skills regarding educational performance, health behavior and delinquency in our study. Further, we find that boys with low noncognitive skills have significantly lower social outcomes compared with girls.

Keywords:

• noncognitive skills
• early childhood
• risk factors
• social outcomes

JEL classifications:

• I20
• J13
• J24

Acknowledgements

We gratefully acknowledge support from the Leibniz Association, Bonn, Germany, under the grant ‘Noncognitive Skills: Acquisition and Economic Consequences.’ Manfred Laucht thanks the German Research Foundation and the Ministry of Science and Research for their support in conducting MARS. For helpful discussions, we thank Andrea Mühlenweg, Maresa Sprietsma and Francois Lainey. All remaining errors are ours.

Notes

1. At the ages of three months and two years, the interrater reliability was measured in a preliminary study of 30 children. Satisfactory interrater agreement was obtained between two raters (3 months: mean κ = 0.68, range = 0.51–0.84; 2 years: mean κ = 0.82, range = 0.52–1.00).

2. A standard alcohol drink consists of 8–12 g alcohol per drink.

3. Data concerning the tobacco and alcohol consumption were assessed via the Substance Use Questionnaire designed by Müller and Abbet (1991) in collaboration with the World Health Institute.

4. MARS is derived from the German title MAnnheimer Risikokinder Studie.

5. The psychosocial risk factors are based on the family adversity index proposed by Rutter and Quinton (1977).

6. Participants with severe handicaps, IQ < 70, MQ (moral quotient) < 70 or neurological disorder were excluded from the original sample.

7. We calculated joint means and standard deviations for both ages, three months and two years.

8. However, agreement exists to consider temperament as something which develops very early and as biologically rooted differences in behavioral tendencies that are relative stable during the lifetime (Pitzer et al. 2007).

9. Successful interventions from the Abecedarion Project, at the very early beginning of a child’s life, are the support of the so‐called home‐school teachers. They provide individual curriculums for each child and interact with the parents every second week (Currie and Blau 2005).

10. A second problem can be reverse causality. This is especially problematic if social outcomes may influence noncognitive skills and vice versa. Since we observe noncognitive skills during early childhood and social outcomes at least six years later, this source of endogeneity is not problematic in this study.

11. Since we observe each social outcome only for two different ages, the first‐differences estimator is equivalent to the within‐estimator.

12. This is due to the fact that social outcomes differ in duration and age when they are assessed for the first time.

13. We re‐estimate all models using linear specifications. The results are similar. They are available from the authors upon request.

14. Calculating differences in IQ scores between t and t – 1 lead to zero within‐variations in our data.

15. In this study, disabled persons are observed until age 11. Thus, for school grades and autonomy, we have observed their social outcomes; however, for reasons of comparison, we exclude all disabled persons from our analysis (three observations).

16. We re‐examined our main estimation for grades in math and German taking into account that a good grade at the highest school track indicates a higher school performance than a good grade at the lowest school track. Yet, for the total (risk) sample, the results are much stronger for the lowest skill category, and become often insignificant for the middle against the highest skill category. The results are not reported here, but are available from the authors upon request.

17. Note that the coefficients are measured less precisely.