The x-axis shows teacher ratings of children’s behavior at age 6 years split into quartiles, in which group 1 contains children with the lowest scores and group 4 contains those with the highest scores, and the y-axis represents mean annual employment earnings at age 33 to 35 years in US dollars. Boxes represent the interquartile range, with the central horizontal line showing the median and the lower and upper horizontal lines the 25th and 75th percentiles, respectively. Whiskers represent 1.5 times the interquartile range. Outliers are suppressed in accordance with Statistics Canada data protection requirements.
eTable 1. Characteristics of Participants With Successful and Unsuccessful Tax Return Linkages
eTable 2. Missing Data Pattern for Baseline Behavior, IQ, and Family Characteristics
eTable 3. Multivariable Regression With Opposition and Aggression as Separate Predictors
eTable 4. Full Sample Multivariable Regression Model With Sex as a Moderator
eTable 5. Correlation Matrix for Child and Family Characteristics and Earnings
eTable 6. Multivariable Regression Tobit Regression Model
Customize your JAMA Network experience by selecting one or more topics from the list below.
Vergunst F, Tremblay RE, Nagin D, et al. Association Between Childhood Behaviors and Adult Employment Earnings in Canada. JAMA Psychiatry. 2019;76(10):1044–1051. doi:10.1001/jamapsychiatry.2019.1326
Are behaviors in kindergarten associated with employment earnings 30 years later, after controlling for child IQ and family background?
In this study of 2850 participants who were followed up for 30 years, inattention at age 6 years was found to be associated with lower annual earnings at age 33 to 35 years, after adjustment for IQ and family adversity. For male participants only, aggression-opposition was associated with lower annual earnings and prosociality was associated with higher annual earnings.
Kindergarten teachers can identify behaviors associated with lower earnings 3 decades later; early monitoring and support for children exhibiting high inattention, aggression-opposition, and low levels of prosocial behaviors could have long-term socioeconomic advantages for those individuals and society.
Specifying the association between childhood behaviors and adult earnings can inform the development of screening tools and preventive interventions to enhance social integration and economic participation.
To test the association between behaviors at age 6 years and employment earnings at age 33 to 35 years.
Design, Setting, and Participants
This study obtained data from the Quebec Longitudinal Study of Kindergarten Children, a population-based sample of boys and girls (n = 3020) born in 1980 or 1981 in Quebec, Canada, and followed up from January 1, 1985, to December 31, 2015. The data included behavioral ratings by kindergarten teachers when the children were aged 5 or 6 years and 2013 to 2015 government tax returns of those same participants at age 33 to 35 years. Data were analyzed from September 2017 to December 2018.
Main Outcomes and Measures
Mixed-effects linear regression models were used to test the associations between teacher-rated inattention, hyperactivity, aggression, opposition, anxiety, and prosociality at age 6 years and reported annual earnings on income tax returns at age 33 to 35 years. Participant IQ and family adversity were adjusted for in the analysis.
The study included 2850 participants, with a mean (SD) age of 35.9 (0.29) years, of whom 1470 (51.6%) were male and 2740 (96.2%) were white. The mean (SD) personal earnings at follow-up were US $33 300 ($27 500) for men and $19 400 ($15 200) for women. A 1-unit increase in inattention score at age 6 years (males mean [SD], 2.47 [2.42] vs females mean [SD], 1.67 [2.07]) was associated with a decrease in annual earnings of $1271.49 (95% CI, −1908.67 to −634.30) for male participants and $924.25 (95% CI, −1424.44 to −425.46) for female participants. A combined aggression-opposition score (males mean [SD] 2.22 [2.52] vs females mean [SD], 1.05 [1.73]) was associated with a reduction in earnings of $699.83 (95% CI, −1262.49 to −137.17) for males only, albeit with an effect size roughly half that of inattention. A 1-unit increase in prosociality score (males mean [SD], 6.12 [4.30] vs females mean [SD], 7.90 [4.56]) was associated with an increase in earnings of $476.75 (95% CI, 181.53-771.96) for male participants only. A 1-SD reduction in inattention score at age 6 years would theoretically restore $3077 in annual earnings for male participants and $1915 for female participants.
Conclusions and Relevance
In this large population-based sample of kindergarten children, behavioral ratings at 5-6 years were associated with employment earnings 3 decades later, independent of a person’s IQ and family background. Inattention and aggression-opposition were associated with lower annual employment earnings, and prosociality with higher earnings but only among male participants; inattention was the only behavioral predictor of income among girls. Early monitoring and support for children demonstrating high inattention and for boys exhibiting high aggression-opposition and low prosocial behaviors could have long-term advantages for those individuals and society.
To what extent is early behavior associated with a child’s future economic circumstances? A series of long-term studies sought to address this question and found that childhood behavioral problems, such as inattention, hyperactivity, aggression, and opposition, are associated with a range of adverse outcomes1-3 that include unemployment, lower earnings, and less wealth in early adulthood.4-9 These outcomes have implications for individuals and society: People with low earnings are more likely to experience welfare dependence, stress, common mental disorders, ill health, and premature mortality.10-14 If early behavioral problems are associated with adverse life outcomes, including lower earnings, then specifying these behaviors is essential to monitoring and supporting children as early as possible.15,16
The association between childhood behaviors and future earnings is not surprising. Children who fight with their peers, are careless in their work, do not attend to instructions, and do not complete assignments are likely to underperform in school17-19 and subsequently in the workplace as adults,4,20 which may be associated with lower earnings. Conversely, children who exhibit prosocial behaviors have been found to have better peer relations,21 fewer adolescent behavioral problems,22 and higher educational attainment,23 which may increase their social and economic capital as well as employment earnings in adulthood. Long-term studies support these results. Inattention, hyperactivity, and antisocial behavior at age 10 years2,4-6 and low self-control between 3 and 11 years of age7,9 are associated with lower earnings and less wealth in early adulthood (age 26 to 36 years, depending on the study).4-8 Socioemotional skills during kindergarten have been linked with employment retention at age 25 years,24 and positive peer relations at age 10 years are associated with higher earnings at age 26 years.6
Early behaviors are modifiable,16,25 arguably more so than traditional factors associated with earnings such as IQ and socioeconomic status, making them key targets for early intervention.26 The long-term studies cited earlier are highly relevant for researchers and policymakers who seek to mitigate the harmful consequences of disruptive behaviors or harness the advantages of early prosocial behaviors, but at least 5 methodologic limitations of these studies temper their value.
First, most previous studies assessed children’s behavioral problems during middle childhood (eg, 10 years of age)4-6 or across multiple years (eg, from 3 to 11 years of age).7,9 This method is not optimal because, by age 10 years, the prospects for successful preventive intervention are diminishing,15 and waiting for several years of behavioral assessments before identification and intervention can occur is impractical. From a cost-effectiveness perspective, behavioral assessments should be made at a single time point and as early as possible when children are most likely to gain an advantage from intervention. Kindergarten is the ideal period because, in many countries, it provides the first opportunity for population-wide assessments that include children from low-income backgrounds who are less likely to attend daycare. Second, several previous studies combined multiple behavioral dimensions (eg, inattention, opposition, and aggression) to form composite indexes.7,9,20 This approach obscures the unique additive or interactive contribution of different behaviors to later earnings. An alternative approach is to include the specific behaviors within a single model so that those associated with the outcome27 can be incorporated into targeted intervention programs, which are more effective and efficient than nontargeted generalized interventions.27,28
Third, previous studies typically used self-reported earnings. This approach is susceptible to biases such as selective dropout (eg, among participants with low socioeconomic status), nondeliberate misreporting (eg, lack of knowledge about the value of income, workplace benefits), under- or overreporting owing to social desirability bias, and deliberate falsification. The use of administrative data (ie, government tax returns) minimizes these biases because these data provide population-wide coverage, account for all sources of income (eg, earnings, benefits, and tax reductions), and are recorded through impartial third-party reporting (eg, employers) and therefore superior to self-reports. Fourth, few previous studies examined the potential advantages of explicitly measured prosocial behaviors, which may be independently associated with higher earnings or may counteract the negative associations of behavioral problems. In addition, to our knowledge, no previous study has examined the association between kindergarten behaviors and employment earnings obtained from tax return records in a large population-based sample that included male and female participants over a follow-up period of 3 decades.
The present study sought to address these limitations of previous research by examining the association between 6 prevalent childhood behaviors assessed in kindergarten (inattention, hyperactivity, aggression, opposition, anxiety, and prosociality) and annual earnings at age 33 to 35 years as measured by tax records. Behavioral ratings at age 5 to 6 years were obtained from kindergarten teachers, and earnings were obtained from government tax return records. A further aim of this study was to estimate the economic burdens of early behavioral problems by calculating the earnings that would be theoretically restored given a 1-SD reduction in a given behavior.
Data for the present study were obtained from the Quebec Longitudinal Study of Kindergarten Children, a population-based sample of boys and girls born in 1980 or 1981 in Quebec, Canada, and followed up from January 1, 1985, to December 31, 2015. Characteristics of this sample (n = 3020) are described elsewhere.29 This study was approved by the University of Montreal Ethics Board and Statistics Canada. Written informed consent was obtained from children’s parents prior to participation.
Behavioral ratings or scores were obtained from teachers, who used the Social Behavior Questionnaire at the end of kindergarten when their students were aged 5 or 6 years.30 The instrument has been validated31,32 and has good predictive validity for a range of outcomes, including conduct problems, adolescent violent and nonviolent delinquency, substance use, and school completion.33-35 Behaviors were assessed as follows: inattention (4 items) included lacking concentration, being easily distracted, having one’s head in the clouds, and lacking persistence; hyperactivity (2 items) included feeling agitated or fidgety and moving constantly; physical aggression (3 items) included fighting, bullying or intimidating, and kicking or biting; opposition (5 items) included disobeying, refusing to share materials, blaming others, being inconsiderate, and being irritable; anxiety (3 items) included being fearful or afraid of new things or situations, worrying about many things, and crying easily; and prosociality (10 items) included stopping quarrels or disputes, inviting bystanders to join in a game, helping someone who has been hurt, showing sympathy, and comforting a child who is crying or upset, among others.36 Behaviors were rated on a 3-point scale, with 0 indicating never or not true; 1, sometimes or somewhat true; and 2, often or very true. The α score for inattention was α = .85; hyperactivity, α = .89; opposition, α = .86; aggression, α = .86; anxiety, α = .74; and prosociality, α = .91.
Outcome data were obtained from government tax returns and linked to the Quebec Longitudinal Study of Kindergarten Children participant records through Statistics Canada.37 Tax return data for each year included personal and family pretax income, marital status, and number of children in the household. Personal earnings were defined as all annual pretax wages, salaries, and commissions, excluding income from capital gains. The 3 most recent available tax returns (2013-2015) at the time of analysis were averaged and used as the main outcome (the Pearson correlation coefficient across the 3 years ranged from 0.93 to 0.85 for males and 0.82 to 0.72 for females). Earnings data were collected as Canadian dollars but are reported as US dollars in this article (Ca $1 = US $0.75).
Cognitive ability was assessed using the Sentence Completion Test38 obtained at age 13 years. This instrument correlates highly with other verbal and nonverbal measures of intelligence and educational achievement across a range of populations.39 Family adversity was calculated by combining the following variables assessed at age 6 years: parental educational level (total years), parental age at birth of the first child, parental occupational status, and family structure (intact vs nonintact), using the criteria developed by Blishen et al.40 Families at or below the 30th percentile for each of these indexes (or nonintact families) were coded as having 1 adversity point. Scores were standardized on a 0 to 1 scale.
Participants with tax return data at age 33 to 35 years (n = 2850 [94.4%]) were retained for analysis (eTable 1 in the Supplement). To account for missing outcome data, inverse probability weights were generated and included for sensitivity analyses in all models (variables of missingness were sex, aggression, and family adversity). After confirming the missing at random assumption, missing data (eTable 2 in the Supplement) used multiple imputations by chained equations.41 Models were estimated across 80 data sets, and the results were pooled.
Mixed-effects linear regression models with robust SEs were used to examine the association between childhood behavior and adult earnings with adjustment for clustering of scores at the regional, school, and classroom levels using clustered standard errors. We controlled for child IQ and family adversity. Aggression and opposition were highly correlated (Pearson r = 0.77), and we combined these variables to create a single variable (results of the uncombined model are shown in eTable 3 in the Supplement). Sex differences in childhood behavioral problems and employment earnings are well documented (male participants in our sample had higher variance in earnings and all childhood behavioral scores, except prosociality), and all analyses were conducted separately for male and female participants (results of the combined model with sex entered as a moderator are shown in eTable 4 in the Supplement). Because behaviors could operate interactively rather than additively, 2-way interactions between all behaviors were examined, with all continuous interaction terms centered prior to analysis. Effect sizes were calculated as the ratio of the coefficient estimate to the SD of the variable. Standardized effect sizes were calculated for the overall models using Cohen f2 (f2 ≥0.02 represented small, f2 ≥0.15 represented medium, and f2 ≥0.35 represented large effect sizes).42
To estimate earnings lost over a 40-year career, we calculated a financial effect (ƒe) as follows:
in which βj was the estimate effect of variable j (eg, inattention) on earnings and stdj was the SD of variable j. The financial effect measured the present value (assuming a discount rate of 3%) over a 40-year work career of a 1-SD improvement in variable j. The direction of improvement depended on the variable: for inattention, financial effect measured the earnings gain from a 1-SD decrease, whereas for prosociality, it measured the earnings gain from a 1-SD increase.
A number of participants (n = 336 [11.7%]) reported 0 earnings for the 2013 to 2015 tax years. To ensure that this analysis was not affected by the presence of outliers at the lower- and upper-income extremes, we used tobit regression analysis, with censoring of earnings below $1000 and above $110 000 for men and above $80 000 for women.
All analyses were conducted with Stata, version 14 (StataCorp LLC). Statistical significance was set at P = .05, and all tests were 2-tailed. Data were analyzed from September 2017 to December 2018.
In total, 2850 participants (1470 [51.6%] male and 2740 [96.2%] white, with a mean [SD] age of 35.9 [0.29] years) were included in the study. Baseline child and family characteristics of participants with complete outcome data are shown in Table 1. Excluded male participants (120 [4.0%]) had higher levels of hyperactivity, aggression, and family adversity at age 6 years, whereas excluded female participants (50 [1.7%]) had higher levels of aggression (eTable 1 in the Supplement).
Participant earnings and family circumstances at follow-up are presented in Table 2. Mean (SD) personal earning at age 33 to 35 years were $33 300 ($27 500) for men and $19 400 ($15 200) for women. Compared with men, women were more likely to be married or cohabiting (860 [61.2%] vs 900 [66.3%]) and to have children living in the household (62.4% vs 76.6%).
Bivariate associations showed that all kindergarten behaviors, as well as IQ and family adversity, were statistically significantly associated with earnings at follow-up (eTable 5 in the Supplement). Multivariable tests of association between behavioral ratings at age 6 years and earnings at age 33 to 35 years, including adjustments for IQ and family adversity, are shown in Table 3. A 1-unit increase in inattention score (males mean [SD], 2.47 [2.42] vs females mean [SD], 1.67 [2.07]) was associated with lower annual earnings for men ($1271.49; 95% CI, −1908.67 to −634.30) and women ($924.25; 95% CI, −1424.44 to −425.46).
Among male participants only, aggression-opposition (males mean [SD] 2.22 [2.52] vs females mean [SD], 1.05 [1.73]) was associated with lower earnings ($699.83; 95% CI, −1262.49 to −137.17), and a 1-unit increase in prosociality (males mean [SD], 6.12 [4.30] vs females mean [SD], 7.90 [4.56]) was associated with higher annual earnings ($476.75; 95% CI, 181.53-771.96). Mean personal earnings associated with these behaviors, split by quartile, are shown in the Figure.
Higher child IQ was associated with higher earnings, and family adversity was associated with lower earnings for both male and female participants. Inclusion of the inverse probability weights to account for missing outcome data did not alter the results.
Standardized effect size (f2) for the overall models was 0.11 for male participants and 0.12 for female participants. For males, effect sizes were largest for inattention followed by aggression-opposition and prosociality, whereas for females the effect size was largest for inattention. No statistically significant interactions were found. In monetary terms, a 1-SD reduction of inattention scores for male participants would theoretically restore $3077 per year, whereas a 1-SD reduction for female participants would restore $1915 per year. Over the course of a 40-year career, the amount restored would total $73 232 for men and $45 569 for women.
Participants who reported 0 earnings (336 [11.7%]) were more likely to have behavioral problems, lower academic school performance, and greater family adversity during childhood. The tobit regression analysis, which accounted for the presence of outliers at the lower- and upper-income extremes, confirmed these findings with effect sizes that were equivalent to those in the main analysis (eTable 6 in the Supplement).
To our knowledge, this study is the first to associate behaviors assessed in kindergarten with annual employment earnings obtained from government tax returns 3 decades later in a large population-based sample of male and female participants. Teacher ratings of inattention at age 6 years seemed to be associated with lower earnings for both sexes, whereas ratings of aggression-opposition were associated with lower earnings for males only, after adjustment for childhood IQ and family adversity, albeit with an effect size roughly half that of inattention; higher prosociality was associated with higher earnings for males.
This study adds to a growing body of literature showing that childhood inattention,4,6 antisocial behavior,4,5 and low levels of prosociality6,24,43 act as channels for adverse social and economic outcomes in adulthood. It confirms that inattention is among the most important early childhood behaviors associated with lower earnings in adulthood, after accounting for other factors.4,6,43 It is 1 of the first studies to our knowledge to associate explicitly measured childhood prosocial behaviors with earnings, and it confirms that prosociality in kindergarten is associated with higher earnings 30 years later in male participants from low-income backgrounds.43 However, this association did not generalize to female participants. The absence of interactions between behaviors is consistent with at least 2 previous studies.4,43
We found that a 1-SD reduction in inattention at age 6 years would be associated with a theoretical gain in earnings of $73 232 for men and $45 569 for women over a 40-year working career. Because low levels of prosociality and combined aggression-opposition were also independently associated with loss of earnings, this monetary amount could be considerably larger for males. In the long term, lost earnings owing to early behavioral problems could accumulate as negative life events are compounded (eg, withdrawal from or underperformance at school, criminal conviction) and as individuals with limited education and skills become trapped in job sectors with little wage growth.
Standardized effect sizes were in the small range, a finding consistent with that of the few previous studies. However, given this study’s focus on specific behaviors (rather than composite indexes7,9,20), and the long duration of follow-up (30 years), these associations may represent conservative estimates. In concrete terms, however, our results show that the long-term losses to earnings associated with early behavior problems are not trivial.
We found no association between prosociality and earnings among female participants. Although previous studies have reported associations between kindergarten socioemotional skills and employment tenure at age 25 years24 as well as between positive peer relations at age 10 years and earnings at age 26 years6 among males and females, these studies differed in several ways, including the age at initial assessment, behaviors assessed, sample sizes, follow-up duration, and lack of control for anxiety, making a comparison to the present study problematic. One explanation for the present result is that the study lacked statistical power to detect an association over the long follow-up period given the use of a single behavioral assessment in kindergarten. Prosocial behaviors were at higher levels for female than for male participants at baseline and should remain relatively stable across development34; this finding suggests that key mediating mechanisms (eg, educational and occupational attainment) may operate differently for men and women. Another possibility, although speculative, was that female participants with prosocial behaviors were more likely to pursue careers in socially oriented but not traditionally high-earning fields (eg, education, health care).
Several mediating pathways may underlie the associations observed in this study. Inattention has been repeatedly associated with low educational attainment, which in turn has been associated with lower occupational status and earnings.44 Antisocial behavior and substance misuse in adolescence, which are more prevalent among individuals with attention45 and behavioral problems,46 have been shown to disrupt educational achievement and harm occupational attainment and work performance.5 Peer rejection has been associated with inattention, aggression, and low levels of prosociality,47 representing another pathway to reduced earnings. An association with delinquent peers could increase the risk of antisocial behavior and low educational attainment. Prosocial behaviors, in contrast, should be associated with better peer or colleague relations,21 educational achievements,23 and employment opportunities, all of which are associated with higher earnings.
The results of this study suggest that, at kindergarten age, inattention was 1 of the behaviors most associated (by effect size) with future earnings, followed by aggression-opposition among male participants, and that screening and preventive interventions in the early years that aim to improve economic participation should target inattention in addition to aggression-opposition. Prosocial behaviors should also be monitored. Screening should begin in kindergarten, in which population-wide assessment is feasible, teachers have a sense of normative social behavior, and teachers’ ratings of children’s behaviors have good predictive validity across a range of outcomes.18,33,48 Support and prevention programs that target inattention28,49 and low levels of prosociality50-52 in elementary school–aged children can improve social integration and educational attainment and thus help close the achievement gap. That family adversity during childhood was associated with future earnings more than any child behavioral variable underscores the importance of reducing childhood adversity, beginning in pregnancy, through comprehensive family support and social engagement efforts.53 These efforts should enhance an individual’s social integration and economic participation in the long term.
This study’s strengths are its prospective design, long duration of follow-up (30 years), and focus on specific behaviors assessed at a single time point in early childhood. It is among the first studies to highlight, in a large population-based sample, the enduring association between boys’ prosocial behaviors and their economic outcomes as men.
Several limitations of this study should also be noted. This study did not account for earnings through the informal economy or for unaccounted accumulation of debt. Because this was an association study, we could not assume causality. Numerous intervening life events may have affected the outcome. The role of mediating mechanisms should be tested in future studies, including preventive randomized clinical trials.
Drawing on a large population-based sample, this study was the first, to our knowledge, to test the association between teacher-rated behaviors in kindergarten and employment earnings in adulthood (as reported in tax returns) among male and female participants in Quebec, Canada. Inattention, aggression-opposition, and prosocial behaviors were associated with future earnings. Children exhibiting these behaviors should be monitored and supported.
Accepted for Publication: April 11, 2019.
Corresponding author: Sylvana M. Côté, PhD, Sainte-Justine Hospital Research Center, 3175, Côte Sainte-Catherine, Étage A, Local A-568, Montréal, QC H3T 1C5, Canada (firstname.lastname@example.org).
Published Online: June 19, 2019. doi:10.1001/jamapsychiatry.2019.1326
Author Contributions: Dr Vergunst had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Vergunst, Tremblay, Nagin, Algan, Côté.
Acquisition, analysis, or interpretation of data: Vergunst, Tremblay, Nagin, Beasley, Park, Galera, Vitaro, Côté.
Drafting of the manuscript: Vergunst, Tremblay, Côté
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Vergunst, Tremblay, Nagin, Park, Vitaro, Côté.
Obtained funding: Vergunst, Tremblay, Algan, Vitaro, Côté.
Administrative, technical, or material support: Tremblay, Vitaro, Côté.
Supervision: Tremblay, Nagin, Algan, Park, Côté.
Conflict of Interest Disclosures: Dr Côté reported receiving grants from the University of Bordeaux and serving as the Initiative of Excellence (IdEx) chair in developmental psychopathology during the conduct of the study. No other disclosures were reported.
Funding/Support: This study was supported by the Quebec Conseil Québécois pour la Recherche Sociale and the Formation de Chercheurs et l'Aide à la Recherche funding agencies; National Health Research and Development Program, Canadian Institutes of Health Research, and Social Sciences and Humanities Research Council funding agencies in Canada; the Molson foundation; the US National Consortium on Violence; and Statistics Canada. Dr Vergunst was funded by a Fonds de Recherche du Québec—Santé postdoctoral fellowship. Dr Algan was funded by grant agreement 647870 from the European Research Council, European Union's Horizon 2020 Research and Innovation Programme. Dr Côté was funded through the Origin Chair of the IdEx University of Bordeaux.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank all participating families for their continued commitment to this study.
Create a personal account or sign in to: