Police orders linked to sharp drop in sexual harm, study finds

Sexual risk orders, civil orders that can restrict people who have never been convicted of a sexual offence, are associated with a large reduction in sexual harm, according to the first empirical assessment of the tactic. Researchers analysing police records in England and Wales found that people subject to an order committed 84.5 percent less sexual harm per month afterwards than before, with the reduction rising to about 93.1 percent among the highest-harm subjects once time spent in prison was taken into account. By contrast, a comparison group of people for whom an order was considered but never obtained saw their measured sexual harm rise by 68 percent, although that increase was not statistically significant.

The study, published in the Cambridge Journal of Evidence-Based Policing, was carried out by Dan Whitten, Eleanor Neyroud and Peter Neyroud, who analysed records from the Police National Computer, the database on which all live sexual risk orders in England and Wales must be recorded. The orders were introduced in March 2015 through an amendment to the Sexual Offences Act 2003, made by the Anti-social Behaviour, Crime and Policing Act 2014. To grant an order, a magistrates’ court must be satisfied on the balance of probabilities that an individual has done an act of a sexual nature and that an order is necessary to protect the public from harm. No criminal conviction is required. Breach of an order, however, is a criminal offence carrying up to five years’ imprisonment, which places the orders in a contested category of measures imposed on the legally innocent at a civil standard of proof but enforced with criminal sanctions.

The researchers drew two samples of order subjects from the Police National Computer. The first comprised 50 orders obtained by the Metropolitan Police Service with start dates between September 2021 and August 2023, a group the authors describe as a more representative picture of how the Metropolitan Police uses the orders. The second comprised 55 subjects whose orders were issued between March 2015 and March 2018, chosen because the longer follow-up period allowed a more longitudinal assessment of harm after the intervention. The authors note that this older population is skewed towards orders of longer duration, possibly reflecting perceptions of elevated risk among police and courts. To build a counterfactual, they used case records from the Metropolitan Police’s Department of Legal Services covering 2015 to 2018, identifying 47 cases in which officers began the process of seeking an order but never completed the application.

To measure harm rather than simply count arrests, the team applied the Cambridge Crime Harm Index, which converts each offence into a score based on the prison sentence starting points set out in sentencing guidelines. Arrests for breach of the order itself were excluded, since the orders aim to prevent sexual harm and breach is not itself a sexual offence. Each arrest for a sexual offence during the relevant period, a decade before the order up to the point data were drawn in 2023, was assigned a harm score, and scores were divided by the number of months to produce a harm-per-month figure before and after the order. For the older sample, the researchers also adjusted for time spent in prison, dividing harm by months at liberty.

Statistical testing using the Wilcoxon signed-rank test showed the reductions were significant. In the 2021 to 2023 Metropolitan Police sample, mean harm per month fell 84.5 percent after the order, a difference the authors report as significant at z = -5.15, p < 0.001. In the 2015 to 2018 sample, mean harm per month at liberty fell 93.1 percent, significant at z = -5.85, p < 0.001. In the counterfactual group, mean sexual harm per month rose 68 percent after the point at which an order could have been obtained, but the Wilcoxon test indicated this was not statistically significant. When the three groups’ trajectories were plotted together, both order populations diverged considerably from the counterfactual.

The team also conducted an interrupted time series analysis on the 2015 to 2018 sample, plotting harm by year from five years before the order to five years after, and modelling the series with autoregressive integrated moving average methods to manage significant autocorrelation. The results indicated a statistically significant reduction in harm immediately after the order was granted, followed by a significant sustained downward trend. The authors argue that this timing matters, because subjects had typically already experienced arrests, investigations and sometimes court proceedings before the order was obtained, yet the drop in harm followed the order itself rather than those earlier contacts with the criminal justice system.

Despite these findings, the study reports that the orders are rarely used and are not systematically aimed at the people causing the most harm. Nationally, only 307 orders were obtained in the two years to 2023, with considerable variation between and within police forces. Among Metropolitan Police subjects from 2021 to 2023, all were male, their ethnicity reflected the London resident population, ages ranged from 18 to 73 with an average in the mid to late thirties, and no orders were obtained for child subjects. Some 78 percent of orders were sought after a failed prosecution, meaning the subject had not been prosecuted or had been acquitted, and in 40 percent of cases the subject had three or more failed prosecutions before an order was sought. Applications rest largely on the professional judgement of individual officers rather than any systematic assessment of harm.

That pattern matters, the authors argue, because harm is heavily concentrated. Rank-ordering Metropolitan Police subjects by the harm caused by alleged sexual offending, using the force’s VAWG100 approach, showed that of the 54 subjects accounting for the top 5 percent of sexual harm in 2022, none subsequently received an order, even though a case-level review suggested 26 were likely eligible. The authors estimate that if orders had been obtained for that small cohort, the harm reduction observed in the two order samples would be equivalent to preventing 62 to 65 rape offences, a projection based on broad assumptions but one they say illustrates the potential of even a rudimentary data-driven targeting method.

The findings carry significant caveats. The design sits at level 3 on the Maryland Scale of Scientific Methods, without random assignment, and the authors are explicit that variables other than the order may explain the differences between groups. The counterfactual cases came from a different time period than the full order sample, and the researchers could not determine for each counterfactual case whether the application was abandoned by police, rejected by a court or simply never completed, though they report that most attrition appears to have resulted from officers not completing application files, often for reasons such as workload or case transfer. The counterfactual group also differed in composition, averaging five years younger and over-representing people of black and mixed heritage. The authors further note that the age-crime curve, the well-established tendency for offending to decline with age, could contribute to the observed drop, since the average order recipient’s age in the mid-thirties coincides with the point at which some research suggests sexual offending may begin to decline independently of any order. Arrest records are also only a proxy for harm, and changes in policing practice, such as those following heightened attention to violence against women and girls, could affect the measure.

Given these limitations, the authors describe the results as promising rather than definitive and call for a randomised trial as the next logical step to strengthen causal inference. They suggest such a trial would be ethically defensible because forces appear to have far more eligible subjects than resources to apply for orders, so a trial would still involve more orders being obtained than at present. Within the limits of the methodology, they conclude that the evidence supports using the orders as a primary tactic against violence against women and girls, with preventative impact potentially maximised by increasing use, actively targeting the highest-harm offenders and considering orders at an earlier stage of a subject’s sexual offending. They also acknowledge that a finding of effectiveness does not settle the deeper jurisprudential debate over how far the state may restrict the rights of the legally innocent, but argue that the evidence now informs one half of that calculation: whether the orders actually prevent the harm they are deemed necessary to prevent.

Subject of Research: An observational study of whether sexual risk orders in England and Wales are associated with reduced sexual harm and how the orders are targeted.

Article Title: Sexual risk orders as a tactic to counter sexual violence against women and girls

Article References: Whitten, D., Neyroud, E., & Neyroud, P. (2025). Sexual risk orders as a tactic to counter sexual violence against women and girls. Cambridge Journal of Evidence-Based Policing, 9(1), Article 1. https://doi.org/10.1007/s41887-024-00097-6

Image Credits: AI Generated

DOI: 10.1007/s41887-024-00097-6

Keywords: sexual risk orders, sexual violence, violence against women and girls, crime harm index, policing, counterfactual comparison, interrupted time series, Metropolitan Police, civil orders, harm reduction, randomised trial, evidence-based policing