Presentation on theme: "Home Office 15 October 2008 - 1 The dynamics of illicit behaviour Analysis of initiation trajectories Steve Pudney."— Presentation transcript:
Home Office 15 October 2008 - 1 The dynamics of illicit behaviour Analysis of initiation trajectories Steve Pudney
Home Office 15 October 2008 - 2 Three issues… Age profiles, cohort differences and onset risk Behavioural trajectories – what are the typical sequences of initiation? What factors influence behavioural trajectories?
Home Office 15 October 2008 - 3 Age profiles The typical pattern of offending or drug use is a steep rise (onset) in the teens, followed by a slower decline in the twenties (desistance) Profiles for types of offending & drug use differ in terms of the peak age of onset & prevalence and rates of rise and decline Three ways of looking at profiles: Prevalence by age Age-prevalence profiles by birth cohort Onset risk (hazard) by age
Home Office 15 October 2008 - 4 Age profiles for prevalence in last year Minor theft: age profiles Cohort effects: shop theft
Home Office 15 October 2008 - 5 Age profiles for prevalence in last year Class A drug use Non-class A drug use
Home Office 15 October 2008 - 6 Drug prevalence in last year by birth cohort Cannabis Cocaine clear downward shift from 1986-8 cohort onwards for cannabis none for cocaine
Home Office 15 October 2008 - 7 Violence/disorder in last year by birth cohort Criminal damage Assault with injury clear downward shift from 1986-8 cohort onwards for assault later shift for criminal damage?
Home Office 15 October 2008 - 8 ASB – juvenile & persistent forms 4 ASB types Cohort-specific profile for graffiti
Home Office 15 October 2008 - 9 Prevalence and hazard rates for drug onset Non-class A drugs Hazard rates Age profiles of prevalence hazard rate or onset risk at age t - the proportion of those who are uninitiated on their t- th birthday, who then commit the offence whilst t years of age
Home Office 15 October 2008 - 10 Age, cohort and onset: summary Evidence of a downward shift in age profiles for many crime categories from the mid- or late-1980s cohorts onwards consistent with long-term downward trend in recorded & BCS crime Strong evidence of a downward shift in age-consumption profiles for cannabis consistent with downward trend in BCS and school surveys since 2004 Age of maximum risk of onset generally precedes peak age of prevalence for drug use, less so for crime e.g.: cannabis: peak prevalence at 19 but peak onset hazard at 16; commercial burglary: 15 for both Target age for policy intervention on drugs is earlier than suggested by crude prevalence figures Behavioural characteristics of age profiles used as a basis for grouping of offences/drug types
Home Office 15 October 2008 - 11 Grouped offences GroupConstituents Prevalence last year Everyday theftTheft from school / work8.9 Other minor theftTheft from car / shop / other5.0 Serious theftTheft of car / burglary / robbery1.7 Assault and damageAssault / criminal damage17.5 Juvenile ASBNoisy/rude behaviour + graffiti17.8 Persistent ASBNuisance to neighbours + racist abuse13.2 VSAVolatile substance abuse0.8 CannabisCannabis + non-class A dealing18.2 Other non-class A drugs Poppers / amphetamines4.5 Recreational class A drugs Ecstasy / cocaine / LSD / class A dealing6.3 Problematic class A drugs Heroin / crack0.3
Home Office 15 October 2008 - 12 Grouped offences: sequences of initiation Activity Sample % (all individuals) with occurrence as first type of illicit activity Sample % as only type of illicit activity School/work theft 8.73.1 Other minor theft9.52.3 Serious theft1.20.4 Assault/damage21.79.5 VSA2.00.6 Cannabis/non-class A drug supply17.69.7 Poppers/amphetamines3.21.3 Recreational class A drugs1.60.8 Other class A drugs0.130.09
Home Office 15 October 2008 - 13 The most common sequences of initiation (includes all with prevalence > 1% in full sample) Sequence type Sample proportion (%) All respondents Age 21 in at least one wave (1) No offending or drugs 45.232.4 (2) Assault/damage 7.94.4 (3) Cannabis/non-class A drug supply 7.610.6 (4) School/work theft 2.22.6 (5) Assault/damage cannabis/non-class A drug supply 2.12.0 (6) Other minor theft 1.3- Cumulated total 66.252.0
Home Office 15 October 2008 - 14 What factors are associated with different behavioural sequences? We look only at onset – i.e. what is the 1 st /2 nd /3 rd etc. type of illicit activity occurring within a young persons development trajectory? But onset is very important – early onset is linked to more serious later harms What factors are associated with the occurrence of different trajectory types? Our analysis: Group sequences into classes Use statistical modelling (multinomial logit) to analyse the sequence data Predict the probability that an individual with given characteristics & background will be observed to have any particular sequence class How does that probability differ for two hypothetical individuals with different characteristics?
Home Office 15 October 2008 - 15 Headline findings so far Significant cohort effects – downward trend in offending/drug use Strong gender effect related to crime (especially violence) Strong effects of local social context Significant impact of clear school policy on discipline Fractured family history has strong effect – especially with re- partnering Parenting style is important After allowing for these factors, there is no significant effect of: Social class Ethnicity Religious affiliation
Home Office 15 October 2008 - 16 Influence of demographic characteristics (all individuals observed at least once when aged 16 or under) No offending or drugs Minor crime only Non- class A drugs only Minor crime & non-classA drugs (any order) Minor crime non-class A class A drugs Non-class A class A drugs Other sequences Predicted prob (average person) 55.4%22.6%10.1%8.2%0.5%0.3%2.8% Effect of changing characteristic by 1 unit: Birth year+3.4 *** -3.6 *** +1.2 * -1.2 ** +0.3 ** +0.0-0.2 Max. observed age -5.7 *** -2.6 *** +3.7 *** +2.6 *** +0.6 *** +0.3 *** +1.1 *** Female+7.1 *** -11.0 *** +6.0 *** +0.0+0.2-1.3 ***
Home Office 15 October 2008 - 17 Influence of social context & school (all individuals observed at least once when aged 16 or under) No offending or drugs Minor crime only Non- class A drugs only Minor crime & non-classA drugs (any order) Minor crime non-class A class A drugs Non- class A class A drugs Other sequences Predicted prob (average person) 55.4%22.6%10.1%8.2%0.5%0.3%2.8% Effect of changing characteristic by 1 unit: Local drug problems (change from 0% perception to 100%) -21.4 *** +4.7 +5.0 +9.2 * +1.2 * +0.7+0.5 Friend / relative in trouble with police -31.7 *** +7.1 ** +8.0 *** +10.0 *** +1.1 *** +0.6 ** +4.9 *** Weak school discipline (scale 0-3) -10.6 *** +1.0+3.7 ** +4.3 *** +0.2 +1.1 **
Home Office 15 October 2008 - 18 Influence of family characteristics (all individuals observed at least once when aged 16 or under) No offending or drugs Minor crime only Non- class A drugs only Minor crime & non-classA drugs (any order) Minor crime non-class A class A drugs Non-class A class A drugs Other sequences Probability (average person) 55.4%22.6%10.1%8.2%0.5%0.3%2.8% Effect of changing characteristic by 1 unit: 1 parent family -6.3 ** +1.6 -0.1+2.2+0.4+0.1+2.1 ** Mother + step-dad -12.8 *** +6.8 * -1.3+5.2 ** +0.1-0.2+2.1 Strong parenting (sacle 0-4) +11.4 *** -1.8-2.9 *** -4.2 *** -0.4 *** -0.2 ** -1.8 *** Liberal parenting (sacle 0-4) +19.5 *** -7.4-4.4-6.8 ** +0.0-0.1-0.8
Home Office 15 October 2008 - 19 Postscript: the value of longitudinal surveys A sequence of interviews gives a much better picture of change over time and the dynamics of behaviour – particularly important for young people who are changing fastest. Observation of behavioural sequences gives more plausible analysis of the direction of causation Dynamic concepts like the risk of onset (aka hazard rate) are crucially important for policy and can only be inferred from longitudinal data Crime, drugs and ASB are developmental outcomes. Repeated interviews + retrospective questions about pre-sample family history identifies the drivers of poor development Longitudinal data can separate the behavioural effects of ageing from behavioural differences between birth cohorts – cross-sections confound them Repeat observations can reveal a lot about data reliability and suggest ways of adjusting for reporting error