When Numbers Replace Context: Reading Scotland’s RSO Data Responsibly
Police Scotland now publishes monthly figures for Registered Sex Offender (RSO) numbers. Behind the headline, two files publish totals by policing division and outward postcode. They reveal real distribution. They also hide the most important question.
These are administrative counts. They do not describe offence type, risk category, victim age group, or how cases were evidenced. Treat them as a map of where supervision occurs, not a full picture of what occurred.
The Transparency Paradox
The Glasgow Times has published reporting citing sex, age, and ethnicity data for Glasgow RSOs. Those demographics are not present in the public Police Scotland datasets.
The disconnect: The public gets raw counts. Journalists can access FOI derived or multi agency material. This creates selective transparency. Numbers are published without the context the public assumes they represent.
Source file for divisional totals: January 2026 Divisional DOCX
The January 2026 snapshot
As of January 15, the postcode dataset indicates how many RSOs are managed in the community versus in custody or hospital. The link below is the direct source.
RSO by postcode January 2026 (XLSX)
Roughly seven in ten are managed in the community. The highest raw divisional concentrations include Glasgow (900), Lanarkshire (718), and Lothians and Borders (503). These are counts, not population adjusted rates. A smaller division can have a higher rate with a lower count.
What the published files genuinely do show
Operational footprint
The numbers indicate where supervision capacity is required. They can help identify where resource pressure is likely to exist.
Geographic clustering
Postcode totals show clustering by area. That is useful for local authority planning, but it is not evidence of relative danger without rates and offence detail.
Trend monitoring
Monthly publication allows trend monitoring over time. Trend interpretation still requires caution, because law, enforcement and reporting can all move the numbers.
The question the dataset cannot answer
When the label registered sex offender appears, many people treat it as a moral conclusion. They assume the underlying allegation has been tested in full, in context, with the relevant evidence available to the defence and heard by the jury.
The register, however, records legal status. It does not describe what evidence was available, what evidence was excluded, or how contested facts were framed. It cannot tell you whether a conviction reflects factual certainty or the limits of the process that produced it.
In Scotland, sections 274 and 275 of the Criminal Procedure (Scotland) Act 1995 can restrict the evidence a jury is permitted to hear. That can include relationship context and prior conduct material that would otherwise go to credibility or narrative coherence. Whatever view one takes of these rules in principle, the dataset is silent on their practical effect.
Put plainly: the figures can tell you how many people are registered and where they are managed. They cannot tell you how many are registered following trials where relevant context was excluded, credibility could not be fully tested, or jurors were asked to decide in an evidentially narrowed frame.
Why this matters for public safety and public trust
Accuracy is a public safety issue. If a system treats error as unthinkable, it does not become more trustworthy. It becomes less accountable.
Public datasets that publish counts without interpretive limits encourage a lazy conclusion. They invite the public to treat the register as a census of guilt, rather than a register of legal outcomes.
If Police Scotland can publish monthly totals, Scotland can also publish the context needed to interpret them properly. That includes clear descriptions of what the dataset is and is not, and equal access to demographic and methodological detail that currently appears selectively through reporting.
A responsible way to read RSO statistics
Use these numbers as an index of supervision footprint, not a proxy for moral certainty. Ask what the dataset can measure, and what it cannot measure. Where context is withheld, treat confidence as an assertion, not a conclusion.