Scottish prison population projections: September 2024

7 Technical annex

Terminology definitions

This section includes explanations of modelling practice and terms to describe the level and range for the projections of the total, sentenced and remand populations.

Microsimulation Scenario definitions

Each new edition of this publication since November 2023 has featured nine different scenario variants. These scenarios use common underpinning data on court case progression and the prison population, from SCTS and SPS respectively. The scenarios are distinct from each other in that they rely on differing assumptions about how future levels of court conclusions and remand arrivals might change.

The microsimulation prison population modelling methodology requires each scenario to be simulated repeatedly dozens of times so that the combined output yields confidence intervals for the total, sentenced and remand population for each scenario. The full projection includes the 95% range intervals for all nine scenarios. These are overlapping in figures 13 and 15, such that it is not possible to easily identify individual scenarios. The values denoting the upper and lower estimates of the populations in the scenarios in table 2 represent the corresponding outer bounds of the 95% range interval for the widest ranging scenarios in the projection.

Court throughput rate scenario definitions

The central court throughput rate scenario variants assume that the conclusions per courtroom will be similar over the next few months to the typical level and range over the year leading up to the 1^st August 2024 data cut off date. To note: the unrepresentative ER period has been excluded (i.e., the period sampled was from 1st July 2023 to 30^th June 2024 for the projections in this edition of the publication). The "high" scenario assumes the average and range of case conclusions per court per day will be around 10% greater than it has been over the same period, and the "low" scenario assumes that the average case throughput per court will be 10% smaller. A rolling sampling period means that for each edition of the projections the model uses data on case conclusion rates from the last year, thereby accounting for the most recent trends in each successive update to the modelling. As an aside, previous iterations of the modelling included additional scenarios where assumptions about the level of conclusions in future at a national level was informed by changes to the number of High Court, Sheriff Solemn and Sheriff Summary trial courtrooms in accordance with revisions to the Court Recovery Programme. These are no longer necessary but will be re-introduced if it is appropriate to represent potential further changes to the Court Recovery Programme.

Remand arrival rate scenario definitions

The distribution of remand arrival rates for scenarios which use central remand arrivals are normally based upon the daily levels for the entire 12 month period up to the data cut off date (in this instance, 1^st Aug 2024). However, due to the singular impact of the ER implemented from late June to late July 2024 a sampling period was selected (1^st July 2023 to 30^th June 2024) which omitted almost all of the ER period. This gives the baseline assumptions for remand arrivals, and the start and end dates progress from one edition of the projections to the next so that the sample period covers a year of typical population flows leading up to the data cut off date. The “low” and “high” remand arrival scenarios use assumptions in line with the most extreme remand inflow levels in recent history. The “low” remand arrivals scenario in the latest edition of the projections uses a range of remand arrivals sampled from the contiguous three month period with the lowest remand arrivals seen in recent history (excluding seasonal periods). Hence the “high” remand arrivals scenarios use remand arrivals sampled from the contiguous 3-month period with the highest remand arrivals in recent history: March to May 2024 (see Table 3). Sampling ranges from a longer period of time can give a more representative sample of what could happen in future. When developing scenarios to replicate rates of arrivals/conclusions in more extreme months it may be appropriate to sample from a single month, but this tends to only support sensitivity analysis.