Hate crime: availability of information recorded by the police in Scotland

2. Investigating the suitability of the Interim Vulnerable Persons Database (IVPD) as a source of reliable information on hate crime recorded by the police

Background

When the Scottish Government last published Official Statistics on Racist Incidents Recorded by the Police in Scotland^[17], we outlined for users our plan to investigate the availability of information on a wider range of hate-related activity - which covered other groups beyond race.

This section includes (i) a description of the process used by police officers to record information on incidents with a hate element, and (ii) the findings of an exercise undertaken jointly by Scottish Government statisticians and Police Scotland to determine whether the IVPD can provide reliable information on hate crime that comes to the attention of the police.

2.1 Process used by police officers to record information on hate crime

When a member of the public contacts the police to report an incident (or if a police officer is witness to an incident) the information is logged on Police Scotland's System for Tasking and Operational Resource Management (STORM) - this is Police Scotland's national command and control system^[18]. Once raised, an assessment of the incident is made and it is given an initial classification - which could change as investigations proceed. One of these classifications indicates whether there is a hate element to the incident.

STORM is largely used for resource allocation purposes. Depending on the information supplied and the outcome of additional enquiries, the incident may result in the creation of one or more crime reports on the relevant crime management system (CMS). It should be noted that several different crime management systems are used by the different divisions of Police Scotland to record and manage crime.

The STORM and crime management systems operate independently, although all crime reports must have the related STORM incident number recorded for cross referral.

When the incident is complete, it is again classified on STORM, and assigned up to six disposal codes - including whether it has a hate element and (or) whether a crime report has been raised.

Should any incident be assessed to have a hate element, a record should be added to Police Scotland's Interim Vulnerable Persons Database (IVPD). The purpose of the IVPD is to ensure that any concerns for the victim, or any other person (subjects of concern), are assessed and the appropriate action taken. The IVPD is Police Scotland's national database for recording all hate-related information, allowing them to enhance understanding of the extent of hate-related activity across the country. It enables identification of repeat victims and offenders and allows for a holistic assessment of wellbeing concerns and needs which influence multi-agency investigations, interventions and support. It should be noted that the IVPD operates independently of all other systems.

Figure 1 shows how a hate-related incident may be processed by Police Scotland. This is an iterative process as each of the systems (STORM, CMS and IVPD) should be updated as any investigation progresses.

Figure 1: Process for recording hate-related incidents

2.2 Using the IVPD to produce information on hate crime

Previously, statistics on racist incidents recorded by the police in Scotland were produced using a labour intensive manual collection process which collected data from police systems across Scotland. This was not a sustainable model for producing wider data on hate-related activity recorded by the police, therefore an alternative process was sought.

Whilst information on hate-related incidents is held on the STORM and CMS systems, it cannot be extracted without a manual review of every record to capture details for key variables (such as victim and perpetrator characteristics). There is the facility to record the STORM number on the IVPD, for cross referral, however this is not mandatory or searchable.

Like the STORM and CMS systems, the IVPD is also designed primarily for operational and administrative purposes, rather than statistical ones. However, the design of the IVPD offers more potential for the extraction of regular high level information that could provide a good indication of the scale of hate crime recorded by the police in Scotland.

2.3 Joint exercise to measure data capture within the IVPD

Before any analysis of hate crime data within the IVPD could be carried out, Scottish Government statisticians and Police Scotland wanted to ensure that the recording process, as described above, was functioning effectively. To investigate this, both parties jointly undertook an exercise to measure the proportion of valid hate-related incidents in STORM which were also recorded in the IVPD. If the activity flagged as hate-related in STORM had also been recorded in the IVPD by the investigating officer (or other related party), then this would confirm that the IVPD has been set up in way that adequately captures the total extent of hate-related incidents recorded by the police.

Stage 1 - Data cleaning and auto-matching

When the police create a new incident record within STORM, a STORM reference number is automatically generated. If a new record is subsequently created in the IVPD, there is a field to manually input the corresponding STORM reference number, allowing the incident to be cross referenced between the two systems.

The first stage of this exercise was to match the reference number of all hate-related incidents recorded in STORM to the matching STORM reference numbers within the IVPD (i.e. to confirm via automatic data matching that hate-related incidents in STORM had also been recorded and marked as hate-related in the IVPD)

A challenge with the existing set up in the IVPD is that the field used to record the STORM reference number is a free text field, which can increase the potential for input errors to occur (for example if part of the reference number is missing, mistyped or in the wrong format). Before carrying out the matching exercise, the data from the STORM reference field of the IVPD was cleaned.

The incident numbers from STORM followed a consistent format (within each division) so the references noted in the IVPD were checked through, making sure they followed the appropriate STORM reference format where possible, to maximise the chance of a match being found.

An auto-match was then carried out to identify which hate-related STORM records for the reporting years of 2014-15, 2015-16 and 2016-17 had a corresponding record in the IVPD.

During this preliminary stage it was observed that there was a very low match rate for incidents that were opened in STORM as hate-related but did not include a final or disposal code indicating that the incident had a hate element. Given this very low rate it was assumed that these incidents in STORM were not hate-related, and therefore the investigating officer would not be required to enter a hate-related record in the IVPD.

In order to check this assumption, a sample of 50 STORM incidents which were initially marked as hate-related but not disposed of in this way, were reviewed - to assess whether they were hate-related. The incident notes for those 50 records confirmed that almost all of them were not hate-related (with only one incident containing a hate element, and as such should have been added to the IVPD).

This confirmed that the vast majority of incidents which were flagged as hate-related when initially recorded but not when closed or disposed of are not hate-related in their nature - and can be excluded from the matching exercise.

The matching exercise found that, across the three years of 2014-15 to 2016-17, 71% of hate-related STORM incidents had a corresponding IVPD record. Given the free-text nature of the field used in the IVPD to capture the STORM reference number, it was possible that a further number of hate-related STORM incidents were also recorded in the IVPD, but had failed to match during stage 1 as a result of discrepancies in how the STORM reference number had been entered in the IVPD. In order to get an estimate of the match rate closer to the true value, a second sample-based check was carried out where individual records were looked up.

Stage 2 - Manual sample check

The second stage of the matching exercise involved taking a sample of STORM incidents that were closed as hate-related, but had no initial match with a record in the IVPD during Stage 1. A manual exercise was undertaken to see if those incidents were in the IVPD after all, but had failed to match due to an issue in how the STORM reference number had been entered in the IVPD. This stage of the process was carried out by analysts from Police Scotland.

A sample of 300 incidents were drawn from unmatched hate-related STORM incidents from the third and fourth quarter of 2016-17 (which covered the latest data available at the time of this exercise). The North East and Highland and Islands divisions of Police Scotland could not be included in the sample as the information required from STORM was not available from a central location at the time of analysis (those two divisions comprised 11% of those incidents closed in STORM as hate-related but not found within the IVPD as part of the first matching exercise in the sample time period).

Using information about each incident, Police Scotland analysts attempted to locate them within the IVPD. They used information obtained from the incident such as the description of events, date, time, location and details about those involved to perform searches within the IVPD.

Of the 300 incidents sampled, 176 were identified through this manual checking approach as having a corresponding IVPD record. Although the remaining incidents could not be located within the IVPD, this does not necessarily mean all of them are not there. There could be a discrepancy in the details recorded on either the STORM incident or the IVPD record, meaning some incidents could not be easily located.

Bringing together Stage 1 and Stage 2

It is important to note that some caution should be exercised when interpreting the results of the above review. This is because the checks carried out at Stages 1 and 2 are not guaranteed to identify a corresponding IVPD incident (where there is one) for every STORM incident selected in the sample. There remains the possibility that some of the remaining STORM records (that didn't match with an IVPD record) could be in the IVPD, but weren't specifically identifiable. Furthermore, this exercise will not identify any cases where details of a hate-related incident have been entered into the IVPD without a corresponding hate-related STORM incident being raised.

When the results from stage 1 and 2 are brought together, they suggest that at least 88% of hate-related incidents recorded by the police, for the second half of 2016-17, were also recorded in the IVPD. This figure combines the match rate from the automatic data linkage of hate-related incidents within STORM and the IVPD (using STORM reference numbers) with the additional matches found using the manual checks outlined in stage 2. For the reasons noted above, this is likely to be a somewhat conservative estimate.

Summary of investigation

Both Scottish Government statisticians and Police Scotland consider that the match rate between the two systems is sufficient, at this stage, to allow summary information to be drawn from the IVPD that provides a good indication of the total volume of hate crime recorded by Police Scotland. This is because the match rate of at least 88% is felt to be relatively high, given the process of entering information into STORM and the IVPD is likely to contain some degree of administrative or input error (given the thousands of officers using these systems on a day-to-day basis). As such it can be used to provide summary information on hate crimes that come to the attention of the police for each of the five hate strands (race, religion, sexual orientation, disability and transgender identity). This information is presented in section 3 of the report.

Whilst these checks provided a good level of assurance, it should be noted the second stage used a sample of records from a relatively short time period (six months). As such there may be merit in running a similar check at some point in the future, as the IVPD continues to develop. The format of STORM reference numbers is now consistent across Police Scotland which may allow the format of the STORM reference field within the IVPD to be fixed, to ensure only a correctly formatted reference can be entered. This change would ensure a higher level of consistency between the systems, and allow a future matching exercise to be carried out in a more efficient manner.

Beyond the summary information, included in section 3, the IVPD also includes more detailed information on the nature and circumstances of each hate crime that comes to the attention of the police (for example the characteristics of the people involved). This also extends to incidents that may have been hate-related but did not constitute a criminal offence. The development of this additional information on the nature of police recorded hate crime is considered further in section 4.