CRIMINAL CAREERS AND POPULATION REGISTERS PROBING THEORETICAL PERSPECTIVES

1
CRIMINAL CAREERS AND POPULATION REGISTERS
PROBING THEORETICAL PERSPECTIVES

• Keith Soothill
• Mogens Nygaard Christoffersen
• Lancaster University
• The Danish National Institute of Social Research

2
CONSIDERING CRIMINAL CAREERS FOR THE TESTING OF
THEORIES

• USING DENMARK AS THE SOCIAL LABORATORY

3
Paradigms relating to perspectives on crime
reductions

• Parental child rearing methods
• Structural factors relating to family during
  adolescence
• Neighbourhoods community crime prevention
• Individual resource deficits

4
POPULATION REGISTERS

• ADVANTAGES
• Focuses on the whole population, not samples
  therefore, large numbers.
• Standard measures.
• Collected (more or less) consistently over time
  thus, can measure changes over time using
  different birth cohorts.
• Avoids recall bias.
• DISADVANTAGES
• Limited to official information which is
  collected by administrators.
• Limited types of information.

5
Table 1 Information selected from the
population-based registers used in the Danish
cohort study

• Population statistics gender, age, marital
  status, address
• Medical register on vital statistics cause of
  death, suicide
• Unemployment statistics branch of trade,
  unemployment
• Education statistics grades
• Educational classification module schooling,
  vocational training
• Social Assistance Act statistics children in
  care
• Integrated Database for Labour
• Market Research occupation, unemployment
• Crime statistics violation, adjudication,
  imprisonment
• Income compensation benefits social benefit,
  duration
• Fertility Database no. of siblings, parity,
  link to parents
• National inpatient register ICD-8 diagnoses
  (somatic), e.g. abortion
• National psychiatric register ICD-8 diagnoses
  (psychiatric)

6
Data for the present study

• All males born in Denmark in 1980 and followed
  until 2003 (N29,944)
• We have yearly data from birth to age 23 on
  convictions and on social factors

7
DOMAINS OF INTEREST

• Social background (5 variables)
• Family background (2 variables)
• Intergenerational transfer (3 variables)
• Educational qualifications of parents (2
  variables)
• Parental employment and poverty (3 variables)
• Disadvantaged area (2 variables)
• Individual resources (10 variables)
• Ethnic background (1 variable)

8
DOMAINS OF INTEREST AND ALLOCATION TO PARADIGMS

• PARENTING
• Social background (5 variables)
• Family background (2 variables)
• Intergenerational transfer (3 variables)
• --------------------------------------------------
  -------------------------------------------------
• FAMILY STRUCTURAL FACTORS
• Educational qualifications of parents (2
  variables)
• Parental employment and poverty (3 variables)
• --------------------------------------------------
  -----------------------------------------------
• NEIGHBOURHOODS
• Disadvantaged area (2 variables)
• --------------------------------------------------
  -----------------------------------------------
• INDIVIDUAL RESOURCE DEFICITS
• Individual resources (10 variables)
• --------------------------------------------------
  -----------------------------------------------
• NOT ALLOCATED
• Ethnic background (1 variable)
• --------------------------------------------------
  -----------------------------------------------

9
PARENTAL CHILD REARING PARADIGM10 variables used

• Social background
• Parental substance abuse
• Parental mental illness
• Domestic violence
• Parental suicidal behaviour
• Battered child syndrome
• Family background
• Child in care (looked after children)
• Family separation
• Intergenerational transfer
• Mother teenager
• Mother convicted
• Father convicted

10
FAMILY STRUCTURAL FACTORS 5 variables used

• Educational qualifications of parents
• Mother has professional qualification
• Father has professional qualification
• Parental employment and poverty
• Parental unemployment gt21 weeks
• Poverty (lt40 of median income)
• Parental disability pension

11
NEIGHBOURHOODS2 variables

• Neighbourhoods
• Disadvantaged area
• Rented housing (not self-owner)

12
INDIVIDUAL RESOURCE DEFICITS10 variables

• Individual resources
• Unemployment gt 21 weeks
• Didnt pass basic schooling level
• Not in process of training or education
• Graduated
• Poverty (lt50 of median level)
• Not living with a parent
• Psychiatric disorder
• Attempted suicide
• Drug abuse
• Alcohol abuse

13
OUTCOME FACTORS

• We look at the first conviction for
• Shoplifting (n1,208)
• Burglary (n1,585)
• Violent offences (n1,778)
• (Different types of offences but the offences are
  those which anyone could engage in unlike, say,
  drunk driving or embezzlement where the person
  needs to have a car or be in work).

14
Method

• As the data is longitudinal and the data is
  measured yearly, we adopt a Cox discrete time
  model to analyse the data.
• The outcome is a yearly (0,1) dichotomous
  variable with 1 for the first conviction for a
  target offence.
• Risk factors are also (0,1) dichotomous variables
  indicating presence or absence of risk at each
  year.
• A year effect is also included in the model.

15
Odds ratios (unadjusted single factors) for the
three offences re PARENTAL FACTORS

• Shoplifting Burglary Violence
• Social background
• Parental substance abuse 2.3 2.9 2.3
• Parental mental illness 2.0 2.5 2.1
• Domestic violence 2.9 4.1 4.1
• Parental suicidal behaviour 3.2 4.4 4.1
• Battered child syndrome 2.6 3.0 2.5
• Family background
• Child in care (looked after children)
  4.1 6.8 4.2
• Family separation 2.8 3.4 2.9
• Intergenerational transfer
• Mother teenager 2.4 3.3 3.0
• Mother convicted 4.2 3.0 4.0
• Father convicted 4.0 3.8 4.1
• NS not significant 0.05 level 0.01 level
  0.001 level

16
Odds ratios (unadjusted single factors) for the
three offences re FAMILY STRUCTURAL AND
NEIGHBOURHOOD FACTORS

• Shoplifting Burglary Violence
• FAMILY STRUCTURAL FACTORS
• Educational qualifications of parents
• Mother has professional qualification
  0.6 0.4 0.4
• Father has professional qualification
  0.5 0.4 0.3
• Parental employment and poverty
• Parental unemployment gt21 weeks
  2.8 3.3 3.1
• Poverty (lt40 of median income)
  1.9 2.6 2.2
• Parental disability pension
  1.9 2.2 2.1
• NEIGHBOURHOODS
• Disadvantaged area 3.3 2.9
  3.5
• Rented housing (not self-owner)
  2.2 2.3 2.3
• NS not significant 0.05 level 0.01 level
  0.001 level

17
Odds ratios (unadjusted single factors) for the
three offences re INDIVIDUAL RESOURCE FACTORS

• Shoplifting Burglary Violence
• Individual resources
• Unemployment gt 21 weeks
  2.8 5.0 3.4
• Didnt pass basic schooling level
  5.0 6.9 5.7
• Not in process of training or education
  1.5 2.2 2.1
• Graduated 0.3 0.2 0.3
• Poverty (lt50 of median level)
  1.9 2.9 2.0
• Not living with a parent
  1.3 1.6 1.6
• Psychiatric disorder 2.8 3.6 3.3
  
• Attempted suicide 6.1 NS 4.6
• Drug abuse 4.9 7.0 4.6
• Alcohol abuse 2.1 2.8 2.8
• NS not significant 0.05 level 0.01 level
  0.001 level

18
Odds ratios (unadjusted single factors) for the
three offences re ETHNIC BACKGROUND

• Shoplifting Burglary Violence
• Ethnic background
• Ethnic minority 3.4 2.7 3.9
• NS not significant 0.05 level 0.01 level
  0.001 level

19
BUT HOW MANY OF THESE 28 VARIABLES ARE USEFUL IN
THE FINAL MODEL?

• SHOPLIFTING 16 variables
• BURGLARY 18 variables
• VIOLENCE 18 variables
• 14 variables are common to all three outcome
  offences

20
STEPWISE FINAL MODEL (ODDS RATIOS) FOR THE THREE
OFFENCES re PARENTAL FACTORS

• Shoplifting Burglary Violence
• Social background
• Parental substance abuse NS NS NS
• Parental mental illness NS NS NS
• Domestic violence 1.3 1.6 1.8
• Parental suicidal behaviour NS NS NS
• Battered child syndrome NS NS NS
• Family background
• Child in care (looked after children)
  1.8 2.1 1.5
• Family separation 1.6 1.5 1.5
• Intergenerational transfer
• Mother teenager NS 1.3 1.3
• Mother convicted NS NS NS
• Father convicted 1.7 NS 1.6
• NS not significant 0.05 level 0.01 level
  0.001 level

21
STEPWISE FINAL MODEL (ODDS RATIOS) FOR THE THREE
OFFENCES re FAMILY STRUCTURAL AND NEIGHBOURHOOD
FACTORS

• Shoplifting Burglary Violence
• FAMILY STRUCTURAL FACTORS
• Educational qualifications of parents
• Mother has professional qualification
  NS 0.7 0.8
• Father has professional qualification
  0.8 0.8 0.6
• Parental employment and poverty
• Parental unemployment gt21 weeks
  1.6 1.5 1.5
• Poverty (lt40 of median income) NS 1.2
  NS
• Parental disability pension NS NS
  NS
• NEIGHBOURHOODS
• Disadvantaged area 1.4 1.3
  1.5
• Rented housing (not self-owner)
  1.3 1.2 1.2
• NS not significant 0.05 level 0.01 level
  0.001 level

22
STEPWISE FINAL MODEL (ODDS RATIOS) FOR THE THREE
OFFENCES re INDIVIDUAL RESOURCE FACTORS

• Shoplifting Burglary Violence
• Individual resources
• Unemployment gt 21 weeks 1.7 2.0 1.3
• Didnt pass basic schooling level
  1.8 1.9 2.0
• Not in process of training or education
  1.2 1.6 1.2
• Graduated 0.6 0.3 0.4
• Poverty (lt50 of median level)
  1.9 2.9 2.0
• Not living with a parent
  1.6 1.8 1.4
• Psychiatric disorder NS NS 1.4
• Attempted suicide NS NS NS
• Drug abuse 2.5 3.0 NS
• Alcohol abuse 1.6 1.8 1.9
• NS not significant 0.05 level 0.01 level
  0.001 level

23
STEPWISE FINAL MODEL (ODDS RATIOS) FOR THE THREE
OFFENCES re ETHNIC BACKGROUND

• Shoplifting Burglary Violence
• Ethnic background
• Ethnic minority 2.0 1.6 2.4
• NS not significant 0.05 level 0.01 level
  0.001 level

24
THEORETICAL PERSPECTIVES ON CRIME REDUCTION
WHAT TO JETTISON?

• All seem to contribute to a potential explanation
  (but note that variables are not explanations
  to explain, one needs to identify the social
  mechanisms which is beyond the scope of this
  presentation).
• BUT WHICH VARIABLES ARE LIKELY TO BE MORE
  IMPORTANT IN DEVELOPING A WAY FORWARD FOR CRIME
  REDUCTION?

25
DEVELOPING A WAY FORWARD FOR CRIME REDUCTION?

• Either consider highly significant variables (but
  these may be only relevant for a small proportion
  of the population)
• Or consider significant variables that affect a
  larger proportion of the population.
• We take the latter approach in this talk.

26
Counterfactual reduction in convictions

• Assume risk factor has causal role in model
• We then eliminate risk factor from cohort (set it
  to zero for all time periods) keeping all other
  background variables unchanged.
• We then re-estimate probability of conviction for
  each case.
• SO HOW MANY OFFENDERS WOULD NO LONGER OFFEND?

27
Taking the most important variable for each
domain of interest as an example

• Shoplifting Burglary Violence
• (n1,778) (n1,208) (n1,585)
• Social background (5)
• RISK Domestic violence 53 72
  111
• Family background (2)
• RISK Family separation 391 265
  333
• Intergenerational transfer (3)
• RISK Mother teenager N/A 36
  48
• Educational qualifications of parents (2)
• RISK Father has no
• professional qualification 248 205
  507

28
Taking the most important variable for each
domain of interest as an example (continued)

• Shoplifting Burglary Violence
• (n1,778) (n1,208) (n1,585)
• Parental unemployment and poverty (3)
• RISK Parental unemployment gt 21 weeks
  497 302 428
• Disadvantaged area (2)
• RISK Rented housing (not self-owner)
  231 132 127
• Individual resources (10)
• RISK Not graduated 693 809 888
• Ethnic background (1)
• RISK Ethnic minority 160 60
  174

29
SO TO CONCLUDE

• No one theory is sovereign all appear to
  contribute risk.
• Highly significant risk factors may affect
  relatively few people and thus have a small
  impact on crime reduction.
• Counterfactual arguments raise general issues
  relating to social deprivation and education,
  which are highly relevant for crime reduction.

30
SO TO CONCLUDE (2)

• Immigration status as a risk factor is
  particularly tricky