Get documents about "UNC Injury Prevention Research Center
Document Sample
How should we advance our
knowledge of risk assessment
for internet sexual offenders?
R. Karl Hanson & Kelly M. Babchishin
Public Safety Canada
R. Karl Hanson, Ph.D.
Senior Research Scientist
Public Safety Canada
Karl.hanson@ps-sp.gc.ca
Global symposium for examining the relationship between online and
offline offenses and preventing the sexual exploitation of children, Chapel
Hill, North Carolina, April 5-7, 2009
Risk Assessment
• Base rate
– The probability that the “typical” member of
the class will fail
• Risk factors
• Risk evaluations
– Combination of risk factors
What is a risk factor?
• Variation in the factor is associated with
variation in the outcome
• Needs to be established empirically
Sexual
Initial Recidivism
Assessment Follow-up
Yes
Any male victims No
Yes
Only female
victims No
Sexual Recidivism Rates (%)
5 years 10 years 15 years
Rapists 14 21 24
Incest offenders 6 9 13
Girl victim child molesters 9 13 16
Boy victim child molesters 23 28 35
Recidivates?
no yes
yes 32 8 40
Single? (20%)
no
53 7 60
(11.7%)
85 15 100
Measures of Association
• Correlation coefficients – r, phi
• Rate ratio, odds ratio
• Standardized mean differences – d
• Percent correct (hits, false alarms)
• ROC curves
Sexual recidivism by marital status (single)
Study Correlation Sample Size
Fitch .31 139
Hanson .22 191
Quinsey .18 178
Proulx .12 373
Frisbie & Dondis .10 1153
Thornton .07 204
Reddon .04 296
Bonta & Hanson .01 316
Mean = .11 (.07 to .15), Q = 9.62, p >.25
Summary of Meta-Analyses of Risk
Factors for Sexual Offenders
• Hanson & Bussiere, 1996, 1998
• Hanson & Morton-Bourgon, 2004, 2005,
2009
• Mann, Hanson, & Thornton, in preparation
Main Risk Factors
• Young Age
• Sexual Deviancy
• Antisocial Orientation
• Problems with Secure Adult Attachment
• Negative Social Influences
Risk Factors
d N (k)
Young age .26 6,969 (21)
Prior sex offences .39 11,294 (29)
Male victims .22 10,294 (19)
Non-contact sex offences .31 10,238 (22)
Diverse sex crimes .20 6,011 (5)
Risk Factors
d N (k)
Sexual pre-occupation .39 1,119 (6)
Deviant sex interests .31 2,769 (16)
Children (PPG) .32 1,278 (10)
Violence .18 1,523 (8)
Other paraphilia .21 477 (4)
Risk Factors
d N (k)
Any prior offences .32 14,800 (31)
Antisocial Orientation .23 23,012 (65)
Never Married .32 2,850 (8)
Emotional congruence .42 419 (3)
with children
Negative social .26 1,736 (7)
influences
Risk Scales for Sexual Offenders
d N (k)
Static-99 .67 20,010 (63)
RRASOR .60 11,031 (34)
Static-2002 .70 3,330 (8)
MnSOST-R .76 4,672 (12)
Risk Matrix –2000 .67 2,755 (10)
How Similar are Online and
Offline Child Molesters?
• Overlap: Same people
• Overlap: Same characteristics
• Recidivism rates
– Risk factors
– Base rates
Studies examining prior contact sex offences among online offenders
Study N % with contact offences Source
Bourke & Hernandez (2009) 155 84.5% (n = 131) Self-reports
Buschman & Bogaerts (in press) 38 55.3% (n = 21) Self-reports (polygraph)
Coward et al. (2009) 128 32.8% (n = 42) Prior arrests/charges
Elliot et al. (2008) 494 10.9% (n = 54) Prior convictions
Galbreath et al. (2002) 39 7.7% (n = 3) Prior charges
Jung & Gulamhusein (2007) 29 13.8% (n = 4) Prior charges
Laulik et al. (2007) 30 6.7% (n = 2) Prior convictions
Neutze et al. (2009) 108 57.4% (n = 62) Self-reports
Quayle & Taylor (2003) 23 47.8% (n=11) Self-reports
Seto et al. (2006) 100 43% (n = 43) Prior charges
Seto & Eke (2005) 201 11.9% (n = 24) Prior charges
Seto & Eke (2008) 301 5% (n = 15) Prior charges/convictions
Sullivan (2007) 215 13% (n = 28) Convictions
Webb et al. (2007) 90 14.4% (n = 13) Convicted and unconvicted
charges/allegations
Wollak et al. (2005) 1,713 11% (n = 188) Prior arrests
Proportion with prior contact sex 1.00
0.90 Official sources
Self-report sour
0.80
0.70
0.60
offence
0.50
0.40
0.30
0.20
0.10
0.00
0 500 1000 1500 2000 2500
Inverse variance
Note. Two outliers (in terms of inverse variance) were removed. Elliot et al. (2008)
and Wollak et al. (2005)
Proportion of online offenders with a history of contact
sex offences
Overall
proportion
(95% CI) Q N k
Overall .16 (.15-.17) 785.27* 3,664 15
Outlier reduced .14 (.12-.15) 211.10* 3,509 14
(Bourke &
Hernandez, 2009)
Official records .12 (.11-.13) 62.16** 3,212 10
Outlier reduced .12 (.11-.13) 23.14* 3,112 9
(Seto et al., 2006)
Self-report .64 (.60-.68) 111.82** 452 5
Outlier reduced .45 (.40-.51) 17.11* 297 4
(Bourke &
Hernandez, 2009)
Proportion of online offenders with a history of contact
sex offences (random versus fixed effects)
Overall proportion
(95% Confidence Intervals)
Random Fixed
Overall .29 (.19-.38) .16 (.15-.17)
Outlier reduced (Bourke & .23 (.28-.19) .14 (.12-.15)
Hernandez, 2009)
Official records .15 (.11-.19) .12 (.11-.13)
Outlier reduced (Seto et .13 (.10-.16) .12 (.11-.13)
al., 2006)
Self-report .56 (.33-.79) .64 (.60-.68)
Outlier reduced (Bourke & .48 (.33-.62) .45 (.40-.51)
Hernandez, 2009)
1.0
0.8
0.6
0.4
0.2
0.0
* d * * d *
d om d om ixe ix ed d om d om ixe ix ed
an F F an F F
R Ran R Ran
Official Self-report
*outlier removed
Overlap: Same People
• 1 out of 8 (13%) caught for both
• 1 out of 2 (50%) admit to both
How do risk factors cause
recidivism?
• Linear Theory
• Dimensional Theory
• Psychological Theory
– E.g., Theory of Planned Behaviour (Ajzen)
Traffic Accidents
Number of Jobs
Sexual Recidivism
Non-contact
sex offences
Traffic Accidents
Lifestyle Impulsivity
Number of Jobs
Sexual Recidivism
Non-contact Deviant Sexual
sexual offences Interests
+ Pedophilia
+ Emotionally close to kids
- Views adult-child sex as wrong Attitude
- Stake in conformity
+ Bad friends Norms Intention
- Prosocial influences
Perceived
+ Knowledge of offending control
+ Self-efficacy
Actual Control
+ Access to victims
- Surveillance
Offence
+ Pedophilia
+ Emotionally close to kids
- Views adult-child sex as wrong Attitude
- Stake in conformity
+ Deviant internet peers Norms Intention
- Prosocial influences
Perceived
+ Knowledge of internet control
+ Self-efficacy
Actual Control
+ Access to victims
- Surveillance
Offence
