The contested nature of risk factor research

1
The contested nature of risk factor research

• Dr Steve Case
• Swansea University

2
The Risk Factor Research debateRisks are
quantifiable, objective, value-fee and scientific
facts with a consistent, predictable
relationship with offending

• Proponents
• Scientific (control, positivist)
• Clinical (objective, treatable)
• Validated replicated
• Practical, atheoretical

• Opponents
• Anti-positivist
• Unethical
• Strengths misrepresented
• Poorly-understood
• Clumsily implemented

3
Methodological paradoxes of RFR

• Simplistic over-simplification
• Factorisation, developmental bias, psychosocial
  reductionism, aggregation, homogenisation,
  imputation
• Definitive indefinity
• Lack of consensus over how to understand risk
  factors, offending and the nature of the risk
  factor-offending relationship (e.g. causal or
  predictive)
• Replicable incomparability
• Replicability does not imply comparability

4
Simplistic over-simplification

• Artefact risk factor research and risk
  factorology
• transformation of individual, personal social
  risk info into factors amenable to
  probabilistic (statistical) calculation
• over-simplifies the risk factor-offending
  relationship
• replication (statistical reliability) over
  validity
• imputation over explanation
• vague, inadequate proxies for putative causal
  processes (OMahony 2008).
• lack of attention to the active human agent
• transforming a dynamic, interactive set of risk
  processes into static relationships and treating
  diverse phenomena (e.g. unemployment, attitudes)
  as if they were equivalent variables (Pitts 2003)

5
Psychosocial reductionism

• the psychogenic antecedents of criminal
  behaviour (Armstrong 2004 103) in
  individualised domains of family, school, peers,
  neighbourhood, lifestyle psychological
• Neglects constructions of risk, socio-structural
  factors (e.g. societal access routes to
  opportunities), social exclusion the impact of
  locally-specific policy formations
• Partial (in the dual sense of limited and biased)
  understanding explanation of youth offending.
• The Cambridge Study in Delinquent Development
• Explanatory RFs at age 8-10 years
  statistically-predict offending at age 14-15
• ASB in childhood, hyperactivity-impulsivity-attent
  ion deficit, low intelligence low school
  achievement, family criminality, family poverty,
  poor parenting

6
The homogenisation of offending

• Offending as a broad homogenous category
• Little exploration of RFs for specific offences
• Offending v Reoffending
• Offending, Crime Justice Survey (Budd et al
  2005)
• Frequent offending 6 different offences in past
  year
• Serious offending any of 6 serious offences
  (vehicle theft, burglary, robbery, theft from
  person, assault resulting in injury, selling
  class A drugs)
• The Home Office Youth Lifestyle Survey (Graham
  Bowling 1995)
• Ever (lifetime offending) or last year (active)

7
Definitive indefinity What is a RF?

• Causal determine or cause offending
• Predictive increase statistical probability of
  offending
• Linear operate on a continuum or scale
• Multiplicative, cumulative or additive - more RFs
  more likely to offend
• Interactive different combinations of RFs may
  exert different effects when experienced together
  
• Overlapping correlated with each other both
  related to offending, but neither having
  temporal precedence
• Correlational
• Multi-stage - increase the likelihood of another
  RF
• Proxy correlated with RFs for offending
• Challenging inoculate against RFs
• Symptomatic the outcomes of offending

8
Causal or predictive risk factors?

• The claim that past behave is the best predictor
  of future behaviour does not mean that past
  behaviour causes future behaviour (Wikstrom, in
  King and Wincup 2008 133)
• Systematic manipulation of independent variables
  control of potentially extraneous variables
  allows scientific researchers to identify 'cause
  and effect relationships
• Lack of detailed understanding of risk factor
  influence on any level, descriptive, exploratory
  or explanatory, other than statistical.
  Causation as regular associations.
• The problem of causation tends to be sidestepped
  in risk-factor research, resulting in a kind of
  black box explanation whereby causal links are
  assumed rather than specified (Porteus 2007
  271-272)

9
Asset Risk assessment in the YJS

• Practitioners must make quantitative judgements
• To what extent are RFs in each domain associated
  with the likelihood of further offending?
• (0 no association, 1 slight or limited
  indirect association, 2 moderate direct or
  indirect association, 3 quite strong
  association, normally direct, 4 very strong,
  clear and direct association)
• Was the issue linked to past offending?
• Is there a direct or indirect link with
  offending?
• Is the link to offending consistent or
  occasional?
• Is the effect on offending likely to be immediate
  or over a longer period?
• Will the issue lead to offending on its own or
  only when other conditions exist?

10
Indefinitive temporality

• Measurement of RFs at time A offending at time
  B, or
• Exposure to risk offending over a set period of
  time (e.g. 12 months)
• Crude, insensitive temporal measures
• Limited attention to the precise timings of
  exposure to risk factors offending behaviour
• Statistical association time-ordering are
  necessary, but not sufficient, to establish
  causation without explanatory mechanisms
  (Wikstrom 2008)
• The Sex Differences in ASB Study (Moffitt et al
  2001)
• Measurement of all (except 5) risk predictors has
  pre-dated the measurement of adolescent
  antisocial behaviour (assessed between the ages
  of 13 and 18).

11
Replicable incomparability

• the most important risk factors are replicable
  over time and place (Farrington 2003 5).
• Aggregated diffs between homogenised groups
  neglects within-individual change, contextual- or
  cultural-specificity
• replicability does not imply commonality
• ASB and Young People (Rutter et al 1998)
• Prospective longitudinal designs - causal
  questions
• Risk mechanisms (causal risk factors) risk
  indicators (factors associated indirectly with
  the causal process).
• The International Self-Reported Delinquency study
  
• Anglo-American, North West European, Southern
  European
• It seems clear that biological, cultural,
  socialization and environmental factors all play
  a role in the prediction of delinquent behaviour
  

12
Conclusion The validity of RFR

• relies inordinately on measuring analysing risk
  as a broadly-phrased, quantitative factor
  aggregated across groups, thus encouraging a
  focus on the replication of statistical
  differences between-groups rather than
  within-individual changes
• dominated by deterministic probabilistic
  developmental understandings of predictive,
  childhood risk factors at the expense of
  alternative more holistic, complex
  explanations
• lacks coherence a clear, well-developed
  understanding of its central concepts, namely the
  definition of risk factors the nature of their
  relationship with offending
• produces findings that are applied uncritically
  over-simplistically by policy makers more
  interested in broad headlines than addressing the
  details research limitations
• has neglected (yet imputed) two crucial issues
  the validity of risk to the real lives of
  different young people explanation of
  relationships between risk youth offending.