Social Psychiatry and Psychiatric Epidemiology

, Volume 47, Issue 9, pp 1391–1400

What is the long-term outcome of boys who steal at age eight? Findings from the Finnish nationwide “From A Boy To A Man” birth cohort study


    • Department of Child PsychiatryUniversity of Turku
  • Sturla Fossum
    • Regional Centre for Child and Adolescent Mental Health, Faculty of MedicineUniversity of Tromsø
  • John A. Rønning
    • Institute of Clinical MedicineUniversity of Tromsø
    • University Hospital of North Norway
  • Henrik Elonheimo
    • Faculty of LawUniversity of Turku
  • Terja Ristkari
    • Department of Child PsychiatryUniversity of Turku
  • Kirsti Kumpulainen
    • Department of Child PsychiatryUniversity of Eastern Finland
  • Tuula Tamminen
    • Department of Child PsychiatryUniversity of Tampere
  • Jorma Piha
    • Department of Child PsychiatryUniversity of Turku
  • Irma Moilanen
    • Department of Child PsychiatryUniversity of Oulu
  • Fredrik Almqvist
    • Department of Child PsychiatryUniversity of Helsinki
Original Paper

DOI: 10.1007/s00127-011-0455-8

Cite this article as:
Sourander, A., Fossum, S., Rønning, J.A. et al. Soc Psychiatry Psychiatr Epidemiol (2012) 47: 1391. doi:10.1007/s00127-011-0455-8



The aim was to study predictive associations between childhood stealing behavior at the of age 8 years with later psychiatric disorders, criminality or suicide attempts and completed suicides up to the age 25 years in a large representative population-based birth cohort.


The sample includes 2,592 Finnish males born in 1981 with information about stealing from both parents and teachers. Information about psychiatric disorders, criminality, suicide attempts requiring hospital admission and completed suicides was gathered from four different Finnish nationwide registries until the study participants were 25 years old.


One out of ten boys had stealing behavior during the previous 12 months. After adjusting for parental education level and conduct problems or hyperactivity (i.e. potential confounds), stealing at eight independently predicted substance use and antisocial personality disorders, and high level of crimes. Stealing was also associated with completed suicide or severe suicide attempt requiring hospital admission. Comorbid stealing and frequent aggression had the strongest predictive association with any psychiatric diagnosis, crime and completed suicide or severe suicide attempt, while stealing without aggression was not associated with any of the negative outcomes.


Stealing accompanied with aggressivity at age eight is predictive of wide range of adversities. However, no increased risk was observed among the group with stealing behaviors but without aggression.




Numerous studies have shown that Conduct Disorder (CD) behaviors predict a wide range of later adversities, including, e.g. psychiatric disorders, crime, substance use and suicide [111]. Most studies aggregate different types of CD behaviors (aggression, destruction of property, theft, violations of rules) into a single global index, as though the nature of the problems is the same. Similarly, in clinical practice, it seems to be common to treat children, who e.g. steal and children who are aggressive as though the nature of their disordered behavior is similar. However, it is documented that the developmental trajectories of different antisocial behaviors may be significantly different [e.g. 12, 13]. For example, the rate of physical aggression decreases with age, whereas theft appears to increase from childhood to adulthood [14]. The developmental trajectories of physical violence and theft during adolescence and early adulthood have been documented to be different and differently related to neurocognitive functioning [12]. Verbal abilities have been shown to be negatively related to physical aggression while they were positively associated with theft [15]. Thus, global indexes of antisocial behaviors may mask the development of antisocial behavior subtypes and putative causal mechanisms. For example, there is some evidence that stealing in childhood may be a more robust predictor of future delinquency than aggression [16]. However, childhood aggressive behavior has been most fully researched, while stealing in childhood has received rather little research attention.

The need to improve the ability to identify and screen children who are at risk for poor antisocial outcomes is great. Furthermore, a child’s stealing behavior frightens, embarrasses and frustrates parents and may be a reason of contacting services. The present study addresses the question of stealing in early childhood and what role it may play as a specific marker for later poor psychosocial outcomes. The data set used is a large nationwide representative population sample with prospective data collection at age eight and in young adulthood. The primary question is whether the isolated antisocial behavior of stealing in early school years serves as a specific indicator for poor outcomes. This information is highly relevant both for planning of preventive programs and early recognition of children at risk as well for clinicians assessing children who are referred because of stealing behavior.

The specific aims of this study were to examine: (1) the prevalence and cross-informant agreement of stealing behavior among boys at age eight; (2) the cross-sectional associations between stealing behavior at age eight and other psychiatric risk factors; (3) the predictive associations between stealing at age eight and late adolescence and early adulthood adversities (i.e. psychiatric disorders, crime, and severe suicidality).

Subjects and methods


This investigation is a part of the nation-wide “From a Boy to a Man” study, a follow-up study included in the Epidemiological Multicentre Child Psychiatric Study in Finland [1723]. The original study sample was drawn from the total population of Finnish children born in 1981 (n = 60,007). The sample consisted of 6,017 children, or 10% of the base population. The 10% sample of the age cohort was drawn by selecting a representative sample of communities according to their degree of urbanization: urban, suburban, rural. In the small communities, all children born in 1981 belonged to the sample, while in the larger cities, a representative subsample of the area based on school districts was drawn from all the school districts. A child registered in the district belonged to the sample even if he attended school outside the district because of a need for, e.g. special education. Of the selected 6,017 children, 5,813 (97%) took part in the study in 1989. Of these 5,813 children, 2,946 were boys. Complete information about stealing at baseline from both parents and teachers, and follow-up data were available from 2,592 subjects (88% of the original sample).

Data collection at baseline was organized through teachers. The children filled in a questionnaire in the class room. The teacher sent the parent questionnaire via the child to the parents, and the parents returned it in a sealed envelope to the teacher. The teacher sent the parent questionnaires in sealed envelopes, the parents’ written consent sheet, teacher questionnaires and child self-reports to the researcher. At the end of the study, the follow-up sheets, study questionnaires and consent forms were returned to the researcher. Information about stealing was thus gathered from parents and teachers independently and confidentially, with no knowledge of each other’s responses.

Participation in the study was voluntary. Informed consent was obtained from the children’s parents at baseline. The combination of information from the questionnaires and nationwide registries was analysed in such a way that the participants could not be identified. The research plan was approved by the Joint Commission on Ethics of Turku University and Turku University Central Hospital.

Methods at age eight


Similar questions focusing on stealing during the previous 12 months at age eight were included in parent and teacher questionnaires, with probe and response items worded as follows: “Does the child steal?”: 1. “no”, 2. “sometimes”, and 3. “often”. Additional items about stealing were also included in the parent questionnaires: “What does he/she steal?” [alternatives: (1) only minor things like pencils, sweets, toys, small amount of money, i.e. pilfering, (2) more valuable things, (3) stealing both minor things and more valuable things]; “Where does he/she steal?” [alternatives: (1) at home, (2) outside home, (3) both at home and outside home]; “With whom he/she steals?” [alternatives: (1) alone, (2) with group of other children, (3) both alone and with other children].

Psychopathology at age eight

The study used a multi-informant design to obtain several perspectives on the child’s emotional and behavioral symptoms in different surroundings. The parents and the teachers completed the Rutter’s parent questionnaire [24] and the teacher-questionnaire [25], respectively. Both scales have been widely used in child psychiatric research [17, 26]. The conduct scale has questions about stealing, lying, aggression, and defiance; the hyperactivity scale has concerns about restlessness, distractibility, inattention, and the emotional scale addresses shyness, withdrawal and anxiety. Because our interest was to examine associations between stealing and other problem behaviors, the “stealing items” were excluded from the original conduct scale. The children themselves filled in the Children’s Depression Inventory (CDI), which measures depressive symptoms [27]. The CDI has shown good validity for assessing depressive symptoms among children [28]. It has 27 items on a scale ranging from 0 to 2 points, but the question concerning suicide was excluded from our version because it was thought that it could upset the children and parents [17]. Parent and teacher information was combined to generate the sum score of conduct, hyperkinetic and emotional scales.

The different childhood psychopathology domains (conduct not including stealing, hyperkinetic, emotional and depressive) were studied separately as categorical and linear variables. To generate easily interpretable measures of psychopathology, results of the four mental health scales were categorized into below or above the 90th percentile. Of note, scoring above these cut-off points has been shown to predict a large number of adversities 10–15 years later [10].

Aggressivity at age eight

Similar questions focusing on aggression during the previous 12 months were included in parent and teacher questionnaires, with probe and response items worded as follows: “The child fights or quarrels often” with alternatives: (1) “does not apply”, (2) “applies somewhat” and (3) “certainly applies”. Information about aggressivity was based on pooled information from parents and teachers. We categorized the subjects into six groups: (1) those who never stole and had no aggressivity, (2) never stole and had occasional aggressivity, (3) never stole and had frequent aggressivity, (4) stole but had no aggressivity, (5) stole and had occasional aggressivity, (6) stole and had frequent aggressivity.

Additional data at age eight

were collected on: (1) parental education level: father’s or mother’s completion of at least 12 years of education (in Finland, compulsory education consists of 9-year comprehensive school after which education can be continued in vocational school or in upper secondary school concentrating on theoretical subjects), (2) family structure: families were classified as intact (two-biological-parent families) or non-intact (other family structure), (3) school performance: teachers reported if the child’s academic performance was (1) better than average, (2) average, (3) poor; alternatives 1 and 2 were pooled.

Outcome variables

Psychiatric disorders at age 18–23

Finnish men born in 1981 received their obligatory military call-up in 1999, providing the opportunity to reach nearly all the boys in the age group. Military service lasting from six to 12 months is obligatory for Finnish males, and they have a medical examination during the spring of the year when they turn 18 years of age. The purpose of the examination is to obtain a preliminary assessment of their fitness class for military service. After the examination, they must attend a call-up between September and November of the same year. At the military call-up, there is a medical examination to assess whether there are any health changes that could change the preliminary fitness class. By March 2004, 80.4% of the men had served the whole period of military service, i.e. 6, 9, or 12 months, whereas 9.5% had been permanently exempt from service, and 6.8% had done civilian service. Thus, 96.4% of the men had carried out their obligatory service period. The cumulative information on psychiatric diagnosis in the present study was based on the military register information including all psychiatric diagnoses from the call-up health examination in autumn 1999, and the military register information at two time points: October 2002 and March 2004. The diagnoses were given at the mental health examination at call-up, during the military service, or at the health examination evaluating the subjects’ fitness for the military. In the register, the information about psychiatric diagnoses is based on the most recent assessment. Therefore, information about possible psychiatric diagnoses is not necessarily the same at different time points. The more severe and chronic psychiatric diagnoses are usually based on consultation with the specialized psychiatric services, whereas usually a less severe diagnosis may be based only on assessment by a general practitioner. Because the general practitioner gets information from the school and health care system, the accuracy of these diagnoses can be considered rather good. The diagnosis of substance abuse is often supported by criminal records, which the military forces receive for all conscripts.

The subject was classified into the “any psychiatric disorder group” if he had a psychiatric diagnosis according to the ICD-10 classification system at the military call-up medical health examination in 1999, or if he had a psychiatric diagnosis according to the information obtained from the Finnish National Military Register in October 2002 or in March 2004 [19]. According to the information pooled from three different time points, subjects were classified into five groups of disorders: anxiety disorder, depressive disorder, antisocial personality disorder, substance abuse disorder, and psychotic disorder (e.g. schizophrenia and schizophreniform psychosis) groups. If the subject had a psychotic disorder at any of the three time points, then he was not classified into any other group. Otherwise, the subject could belong to more than one disorder group.

Criminal offenses at age 16–20

Data on the cohort’s criminal behavior were gathered through the Finnish National Police Register. This is a nationwide electronic database kept by the administration of the Finnish Police. Access to the register was granted by the Police Department, Ministry of the Interior. A nationwide police register was created after the reform of county administration in 1997. The register includes all suspected offenders caught by the police. However, mere admonitions are not usually covered by the register. Furthermore, municipal parking fines are not included in the register. The present study is limited to acts registered during the years 1998–2001. Data are removed from the police register according to a certain schedule, pertaining to the limitation of prosecution by lapse of time. Data were collected from the register at two time points (in the beginning of the years 2000 and 2002) to ensure that the information concerning offenses during the years 1998–2001 is complete. Register information for the year 1997 was not included because of missing data. Minor traffic offenses have been excluded from the analysis as trivial. Within the four-year period, there were 3,052 registered (other than minor traffic) offenses.

According to information obtained from the police register, subjects were classified into four groups: those who had (1) no registered offenses during the four-year period; (2) one or two offenses; (3) three to five offenses; and (4) more than five offenses, indicating a high level of criminal acts [20]. To study different crime types, criminality was divided into five categories: drug, violent, property, and traffic offenses and drunk driving. In this classification, the subject could belong to more than one offense group. The drug offenses refer to various kinds of drug-related activities: producing, importing, exporting, delivering, selling, purchasing, or merely possessing illegal drugs, which are forbidden in Finland. Violence offenses refer to overt aggressive behavior toward another human being. The main subgroups are various kinds of assaults, battery, and robbery. Property crime included covert behavior targeted not at human beings but at property. This category includes different kinds of stealing, illicit use of a motor vehicle, receiving stolen goods, and vandalism. Economic crime (fraud, embezzlement, and various kinds of forgery) was also included in this category. Traffic offenses consist of reckless driving of various degrees and driving without a license. As mentioned above, minor traffic violations were ignored. Finally, drunk driving offenses presume a blood alcohol concentration >0.05%.

Completed suicides and severe suicide attempts

Information about deaths and the causes of death of cohort members before the end of 2005 were collected from Statistics Finland, and further ascertained from death certificates. Statistics Finland produces statistics on causes of death and on the development of mortality. It also maintains an archive of death certificates from which information or copies of death certificates can be obtained for research purposes prescribed by law. The statistics on causes of death are compiled from data obtained from death certificates, which are supplemented with data from the population information system of the Population Register Center. The statistics on causes of death cover persons who have died in Finland or abroad during the calendar year and who, at the time of death, were domiciled in Finland [29]. The causes of death were coded according to the International Classification of Diseases, ninth revision (ICD-9) before 1996 and according to the International Classification of Diseases, tenth revision (ICD-10), from 1996 onward. Death certificates are completed by pathologists. These certificates include additional information about cause of death and manner of death. For example, in fatal poisonings, the most important toxicological finding is indicated in the death certificate by a code stating the underlying cause of death.

The Finnish Hospital Discharge Register was used to identify all subjects who had a hospital admission with a diagnosis of suicide attempt during the years 1994–2005. The computerized discharge register includes, among other things, data on the date of all hospital admissions, discharge diagnoses, and types of accidental injuries. The Finnish Hospital Register was established in 1967 and its good validity is widely documented in the field of epidemiological research [30, 31]. As reported earlier [11], there was 13 completed suicides and additionally 17 males who were admitted to hospital treatment because of a suicide attempt. Three of the males who were admitted to hospital treatment after a suicide attempt later committed suicide. Therefore, before age 25, 27 males either committed suicide or made a severe suicide attempt requiring hospital treatment.

Statistical analysis

The statistical significance of risk factors on outcome variables was tested with logistic regression analysis. Odds ratios (OR) and 95% confidence intervals (95% CI) were estimated using logistic regression. The multivariate analyses of the categorical predictors for number of criminal offenses were carried out by applying multinomial logistic regression analysis. The multinomial logistic regression analysis for a polychotomous response variable is a generalization of the methodology of logistic regression analysis for a dichotomous response variable [32]. A set of indicator variables described psychiatric status at age eight; the group of children who were negative on each of the screens at age eight was defined as the reference group. Statistical computations were done with SAS System for Windows, release 9.1/2006.


As shown in Table 1, according to parent reports, 7.9% and to teacher reports, 3.1% of boys stole at least sometimes. However, very few boys (0.5%) stole frequently according to parent or teacher reports. The agreement between parent and teacher reports of stealing behaviour was very low (kappa coefficient 0.15, p < 0.001).
Table 1

Parent and teacher agreement on reports of child’s stealing at age eight

Parent report

Teacher report


No n (%)

Sometimes n (%)

Often n (%)


2,338 (90.2)

39 (1.5)

10 (0.4)


170 (6.6)

19 (0.7)

14 (0.5)


2 (0.1)

0 (0)

0 (0)

Kappa coefficient 0.15

p < 0.001

Only parent reports included specific questions about stealing. Almost all (99.0%) stole only minor things (pencils, sweets, toys, small amount of money). Furthermore, 64.8% stole at home only, 26.1% outside home only, and 9.0% both at home and outside home. Furthermore, 51.8% stole alone, 23.6% stole with other children, and 24.1% stole both alone and with other children.

Table 2 shows the cross-sectional associations between stealing, and family background and psychopathology at age eight. In univariate analyses, living in a nonintact family, low parental education level, poor school achievement, conduct problems (not including items about stealing), hyperactivity, and emotional problems (mostly anxiety) were associated with stealing. In multivariate analyses including all explanatory variables shown in Table 2, family structure, conduct and emotional problems were independently associated with stealing. Although the association between other conduct problems and stealing was very strong, the majority (62.9%) of those who stole did not belong to the group scoring above the 90th percentile on the conduct scale. When psychopathology domains were studied as linear variables, the results remained the same.
Table 2

Cross-sectional associations between psychopathology and family background, and stealing at age eight. Results of logistic regression analyses


Stealing at age 8

Univariate analyses

Multivariate analyses

n (%)

OR (95% CI)

OR (95% CI)

Parent education level

 Upper secondary (n = 904)

73 (8.1)


 Lower (n = 1,628)

170 (10.4)

1.3 (1.01–1.7)*

1.0 (0.7–1.3)


 Two biological parents (n = 2,142)

179 (8.4)


 Other (n = 425)

70 (16.5)

2.2 (1.6–2.9)***

1.5 (1.1–2.1)*

School achievement

 Good/moderate (n = 2,174)

197 (9.1)


 Poor (n = 405)

54 (13.3)

1.5 (1.1–2.1)**

1.2 (0.8–1.7)

Conduct scale

 <90 percentile (n = 2,308)

158 (6.9)


 ≥90 percentile (n = 233)

93 (39.9)

9.0 (6.6–12.3)***

8.9 (5.9–13.4)***

Hyperkinetic scale

 <90 percentile (n = 2,324)

192 (8.3)


 ≥90 percentile (n = 243)

60 (24.7)

3.6 (2.6–5.1)***

0.9 (0.6–1.4)

Emotional scale

 <90 percentile (n = 2,259)

203 (9.0)


 ≥90 percentile (n = 320)

50 (15.6)

1.9 (1.3–2.6)***

1.6 (1.1–2.3)*

Depression scale

 <90 percentile (n = 2,267)

212 (9.4)


 ≥90 percentile (n = 295)

38 (12.9)

1.4 (0.99–2.1)

0.8 (0.5–1.3)

OR odds ratio, CI confidence interval

* p < 0.05, ** p < 0.01,*** p < 0.001

Table 3 shows the predictive associations between stealing and a wide range of adverse outcomes 10–17 years later, including psychiatric diagnoses, crime and suicide/serious suicide attempts. Among those who had stolen, 21.7% had a psychiatric diagnoses, 36.6% a crime, 2.8% had committed suicide or been treated at hospital after a suicide attempt, and 44.3% had at least one of these outcomes. When adjusted with parent education level (Model 1), stealing predicted depression, anxiety, substance use, and antisocial personality disorders. Furthermore, all crime types were predicted by stealing. An especially strong association was found between stealing and a high level of repetition of crimes defined as five or more offenses during the 4-year period. When the analyses were adjusted with parent education and conduct symptoms (sum score of parent and teacher reports of conduct symptoms excluding items about stealing) remained associated with substance use and antisocial personality disorder, and drug and property offenses (Model 2). However, when the analyses were adjusted with parental education level and hyperactivity symptoms (sum score of parent and teacher reports of hyperactivity symptoms), the results remained rather similar to those in Model 1.
Table 3

Stealing at age eight, and psychiatric and crime outcomes


Stealing at age 8

Model 1 (adjusted with parent education level)

Model 2 (adjusted with parent education level and conduct symptoms)

Model 3 (adjusted with parent education level and hyperactivity symptoms)

n (%)

OR (95%CI)

OR (95%CI)

OR (95%CI)

Psychiatric diagnoses at age 18–23

Any psychiatric diagnosis

 No (n = 2,313)

198 (78.3)


 Yes (n = 269)

55 (21.7)

2.8 (2.0–3.9)***

1.8 (1.2–2.6)**

2.1 (1.5–3.0)***

Depression (n = 45)

9 (4.4)

2.8 (1.3–5.9)**

2.3 (0.9–5.5)

2.6 (1.2–5.8)*

Anxiety (n = 58)

10 (4.8)

2.3 (1.1–4.6)*

1.4 (0.6–3.2)

1.9 (0.9–4.0)

Substance use (n = 45)

14 (6.6)

5.0 (2.6–9.6)***

2.8 (1.3–6.2)*

3.5 (1.7–7.0)***

Antisocial personality (n = 67)

20 (9.2)

4.6 (2.7–8.1)***

3.1 (1.6–6.0)***

3.4 (1.8–6.1)***

Psychosis (n = 14)

2 (1.0)

1.9 (0.4–8.5)

0.5 (0.1–2.8)

1.2 (0.3–5.7)

Crime at age 16–20

No crime (n = 2,024)

161 (63.4)


1–2 crimes (n = 368)

47 (18.5)

1.8 (1.3–2.5)***

1.2 (0.8–1.8)

1.5 (1.1–2.1)***

3–5 crimes (n = 109)

17 (6.7)

2.1 (1.2–3.6)***

1.1 (0.6–2.0)

1.6 (0.8–2.8)

> 5 crimes (n = 91)

29 (11.4)

5.4 (3.4–8.8)***

3.0 (1.7–5.2)***

3.8 (2.3–6.4)*

Crime types

Drug offense (n = 88)

21 (11.7)

3.7 (2.2–6.3)***

2.4 (1.4–4.4)**

2.8 (1.6–4.8)***

Violent (n = 174)

37 (19.0)

3.1 (2.1–4.7)***

1.6 (0.99–2.6)

2.2 (1.4–3.4)***

Property (n = 270)

55 (25.8)

3.0 (2.1–4.3)***

1.7 (1.1–2.5)*

2.2 (1.5–3.2)***

Traffic (n = 271)

44 (21.8)

2.3 (1.6–3.3)***

1.3 (0.9–2.0)

1.8 (1.2–2.6)**

Drunk driving (n = 131)

23 (12.7)

2.5 (1.5–4.1)***

1.2 (0.7–2.1)

1.9 (1.1–3.1)*

Suicide/severe suicide attempt

No (n = 2,566)

247 (97.2)


Yes (n = 26)

7 (2.8)

3.6 (1.2–8.9)**

1.1 (0.3–3.6)

2.5 (0.99–6.4)

Psychiatric diagnosis or crime or suicide/severe suicide attempt


141 (55.7)



112 (44.3)

2.4 (1.8–3.1)***

1.5 (1.1–2.0)*

1.8 (1.4–2.5)***

Results of logistic regression analyses

OR odds ratio, CI confidence interval

* p < 0.05, ** p < 0.01, *** p < 0.001

Finally, we examined the predictive association between stealing with and without aggressive behavior, and major outcomes (i.e. having psychiatric disorder, crime, completed suicide or severe suicide attempt, or any of these three major outcomes). Information about aggressivity was based on pooled information from parents and teachers. We categorized the subjects into six groups, those who never stole/stole and had no aggressivity, aggressivity sometimes, and frequent aggressivity. As shown in Table 4, boys who both stole and had frequent aggressive behavior were at high risk of all major outcomes.To illustrate this, 26.2% of these boys had a psychiatric disorder, 53.0% a crime, 6.2% had committed suicide or made a severe suicide attempt, and 60.0% had one of these outcomes at follow-up. The figures for boys who did not steal and did not have aggressive behavior (reference group) were 7.0, 15.5, 0.4, and 20.1%, respectively. Among those with stealing and sometimes aggressive status, and no stealing and frequent aggressive status, about 44.4% had at least one adverse outcome. Of boys with no stealing and sometimes aggressive status, 29.9% had adverse outcome. Stealing without aggressivity did not predict any outcomes.
Table 4

Comorbid stealing and aggression at age 8, and adverse outcomes in late adolescence and early adulthood







Psychiatric diagnoses



Psychiatric diagnosis or crime or suicide/severe suicide attempt


n (%)

OR (95%CI)

n (%)

OR (95%CI)

n (%)

OR (95%CI)

n (%)

OR (95%CI)

Never stealing, never aggressive (n = 1,184)

83 (7.0)


183 (15.5)


5 (0.4)


237 (20.1)


Never stealing, sometimes aggressive (n = 1,000)

99 (9.9)

1.5 (1.1–2.0)*

241 (24.1)

1.7 (1.4–2.2)***

10 (1.0)

2.4 (0.8–7.0)

290 (29.9)

1.6 (1.3–2.0)***

Never stealing, frequently aggressive (n = 147)

32 (21.9)

3.7 (2.4–5.8)***

50 (34.0)

2.8 (1.9–4.1)***

4 (2.7)

6.6 (1.8–24.8)**

66 (45.2)

3.3 (2.3–4.7)***

Stealing, never aggressive (n = 52)

7 (13.5)

2.1 (0.9–4.7)

9 (17.3)

1.1 (0.5–2.4)

1 (1.9)

4.6 (0.5–40.3)

13 (25.0)

1.3 (0.7–2.5)

Stealing, sometimes aggressive (n = 135)

31 (23.0)

3.9 (2.5–6.2)***

49 (36.3 )

3.1 (2.1–4.6)***

2 (1.5)

3.5 (0.7–18.5)

60 (44.4)

3.2 (2.2–4.6)***

Stealing, frequently aggressive (n = 66)

17 (26.2)

4.7 (2.6–8.5)***

35 (53.0)

6.2 (3.7–10.3)***

4 (6.1)

15.2 (4.0–58.0)***

39 (60.0)

6.0 (3.6–10.0)***

Results of logistic regression analyses

OR odds ratio, CI confidence interval

p < 0.05, ** p < 0.01, *** p < 0.001

When the outcomes of different groups were compared, having any adversity (psychiatric disorder, committing a criminal offense or completed suicide/severe suicide attempt) was selected as a global indicator for negative outcome. In paired comparisons, stealing and frequent aggressive group membership was a significantly stronger predictor of this outcome than stealing and sometimes aggressive (OR 1.9; 95% CI 1.1–3.4), stealing but not aggressive (OR 1.8; 95% CI 2.0–10.0), no stealing but frequent aggressive (OR 1.8; 95% CI 1.1–3.3), no stealing but sometimes aggressive (OR 3.7; 95% CI 2.2–6.1) or no stealing and not aggressive (OR 6.0; 95% CI 3.6–10.0) status.


One out of 10 boys had stealing behavior, indicating that it is relatively common already at age eight. Stealing was strongly associated with other conduct problems cross-sectionally. This is expected since stealing is one criterion of CD in the Diagnostic and Statistical Manual for Mental Disorders, Fourth Edition (DSM-IV). However, almost two out of three boys who stole did not have a high level of conduct problems, i.e. did not score above the 90th percentile on the conduct scale. The associations between stealing and poor school achievement and other psychiatric problems were low or moderate. Living in other than a two-biological-parent family at age eight was independently associated with stealing. This association may be explained by other risk factors associated with a broken family, e.g. parental mental health problems, substance abuse, family violence, weakened parental support, and parent–child discord. Stealing occurred in almost all cases only occasionally. Very few boys were identified as stealing according to both parents and teachers. Accordingly, several studies have examined cross-informant correlations regarding childhood psychopathology, and have uncovered low agreement between different informants [3335]. Boys usually stole rather minor objects and stealing usually took place at home. The overall picture which emerged from these results is that stealing at age eight seems to be often relatively harmless behavior.

Is stealing in childhood a gateway to future delinquency? Present study suggests that risks for antisocial personality and substance use are elevated among those who steal even after controlling for other conduct problems. Of those who stole, 7% had a substance use disorder, and 9% an antisocial personality disorder, while the figures among those who did not steal at age eight were 1% and 2%, respectively. The variable-centered approach suggests that stealing behavior at age eight may be an early marker for particularly to antisocial personality disorder and substance use disorder, and not only because it is a consequence of other externalizing problems. In other words, stealing among boys at age eight, although the behavior per se seems to be rather harmless, may have an independent long-term effect that goes beyond other externalizing problems.

When criminal offenses were studied, the stealing behavior predicted high repetition of crimes, i.e. having more than five arrests during the four-year period at age 16–20 years, after controlling for other conduct problems. To illustrate this association, 11% of those who stole at age eight had more than five crime offenses, compared with only 3% of those who did not steal. When the analyses were adjusted for hyperactivity problems, associations between stealing, and repeated crimes and all specific crime types remained significant. The findings correspond to those of a previous study indicating that hyperactivity in kindergarten is not a very good predictor of juvenile delinquency, whereas high levels of physical aggression and opposition are [36]. Accordingly, previous studies have shown that stealing in childhood is a marker of risk for delinquent outcomes [37].

A novel finding is that stealing behavior predicted completed suicides or severe suicide attempts. Of those who stole at age eight, 3% either committed suicide or were admitted to hospital after a suicide attempt, while the respective figure for those who did not steal was 1%. These results thus enhance our previous findings that completed suicides and severe suicide attempts in adolescence and early adulthood are predicted by childhood psychopathology [11]. Previous research has shown that impulse control problems and aggressive symptoms are associated with later completed suicides, suicidal acts and ideations [1, 11, 38]. Furthermore, symptom clusters not fulfilling the diagnostic criteria of CD are common among child and adolescent suicides [39]. Previous research has suggested that one of the main targets of effective prevention of youth suicides is to reduce suicide risk factors [40]. Therefore, our finding that stealing at age eight is associated with later severe suicidality need to be replicated.

Using person-centered approach, the subjects were classified into different groups based on stealing (yes or no) and aggression (no, sometimes, frequently). The results show the importance of the comorbidity of aggression and stealing behavior for later psychosocial outcomes. Only 21% of boys who stole had no comorbid aggression, while 25% of boys who stole had frequent aggression. Comorbid stealing and frequent aggression had the strongest predictive association with any psychiatric diagnosis, crime and completed suicide or severe suicide attempt when the reference group was those who did not steal and were not aggressive. A combination of stealing and frequent aggression seemed to potentiate the odds for adversities. For example, those who had frequent aggression and stealing, 60% had at least one negative outcome, compared with 45% of those who were only frequently aggressive. Stealing without aggression was not associated with any of the negative outcomes. Similarly, previous cross-sectional studies have shown that psychiatric problems among boys with both aggression and stealing is significantly more severe as compared to boys only engaging in one of these activities [41]. The results in our study suggest that stealing in childhood combined with aggression may represent the beginning of an antisocial trajectory leading to a wide range of adversities in adulthood. However, boys with stealing behavior at age eight without comorbid aggressivity are not at increased risk for psychiatric disorders or criminality.

The study has several strengths: a nationwide large sample, combining information about childhood psychopathology based on validated and multi-informant measurements, a low attrition rate, and use of national registers. Information about childhood family and other environmental factors was lacking (e.g. parental psychopathology or criminality, family environment and parenting practices, peer influence in adolescence). A further limitation of the study is that information about stealing in childhood was available only at age eight. Information about stealing was based on parent and teacher reports. In adolescence, young men have higher prevalence of stealing in self-reports than in parent reports [35]. Therefore, the prevalence of stealing in childhood may be underestimated.

Although the reliance on diagnoses from military records yielded high follow-up rates, the diagnoses were unstandardized and potentially subject to validity problems. Ideally, psychiatric research diagnoses should be based on carefully monitored assessment procedures (e.g. structured diagnostic interviews). Although this limitation may have affected our findings related to specific psychiatric diagnoses, our overall findings do appear robust across global categories (psychiatric caseness). Given the restrictions on the use of registry data in general, information based on the Finnish national military register gives comprehensive and unique information not available in most countries.

In research about crime a key question is how to gather comprehensive data on crime. While court data would be limited to convictions, the police register includes all criminal offenses that have come to the police’s notice. Corruption in Finland is among the lowest in the world ( and the importance of accurate registering of every offense is emphasized in the instructions given by the Police Administration. Furthermore, the problem of ‘‘false positives’’ is not considered a critical source of bias either, because of the strict regulation and control of the Finnish police. Therefore, the Finnish National Police Registry yields valuable and unique population-based information on crime.

Suicide completers and those suicide attempters requiring hospital admission were pooled together for statistical analysis. Since suicide is a rare phenomenon, even a large sample such as ours was not large enough to perform the statistical analysis separately for these two groups. Post hoc comparison of these two groups among males revealed no significant differences between risk factors at age eight, although the limitations of power in the analysis are obvious. In addition, previous research shows that young suicide completers and young serious suicide attempters are very similar [42].


The purpose to considering utility of isolating one behavior from heterogeneous pool of indicators of antisocial behavior is to try to simplify the process of identifying those who would be most at risk for developing undesirable outcomes so that prevention resources may be used efficiently. The present study shows that stealing is rather common behavior among boys in early school years. Stealing accompanied with frequent aggressivity at age eight is highly predictive of a wide range of adversities, and effective identification and treatment of children with such patterns may reduce e.g. later crime. However, no increased risk was observed among the group with stealing behaviors but without aggression. Therefore, stealing behavior on its own does not appear to be a strong marker for later risk, and may result in unacceptable rates of false positives if taken in isolation.

Copyright information

© Springer-Verlag 2011