Scottish prison population projections

2 Background

Up to March 2020, the Scottish Government produced short-term prison population projections using time-series based forecasting methods applied to prison population trends. These projections were based on ARIMA (Autoregressive Integrated Moving Average) models, which are commonly used in other European jurisdictions. ARIMA projections are based on historical trends in the average monthly prison population and indicate the likely trajectory of the population size should trends in crime and justice remain stable. As ARIMA forecasting extrapolates from historical data, it cannot model changes to the prison population that are not already reflected in past trends. Hence, following the rapid large-scale changes which occurred in the criminal justice system during the Covid-19 pandemic, ARIMA modelling was no longer considered appropriate for short term prison population projections.

The pandemic has had profound impacts on justice system activity (e.g., court capacity and throughput) and the flows (e.g., remand arrivals and departures, sentence arrivals and departures) impacting on the prison population. Therefore, the Scottish Government has recognised the need to develop alternative projection models that take into account how the system activity and flows that drive the prison population are likely to change, and use these to estimate short-term changes in the population. To that end, in consultation with justice partners, the development of microsimulation scenario modelling of the prison population has been undertaken since 2021. This approach is more appropriate for modelling the prison population post-pandemic and through the initial recovery period.

The 'microsimulation' scenario modelling adopted simulates arrivals and departures of individuals to and from the prison population and then estimates the number of individuals in prison on a particular date in the future. The microsimulation model uses individual-level data to construct a representation of a target population. The model can simulate, using known probabilities, changes in individuals' status over time.

To produce outputs, the model draws on particular court activity data, for example, levels of court hearings at different stages (pre-trial / trial) in summary and solemn courts. This data is also used to estimate levels of remand arrivals in the coming months. The model includes assumptions on future court capacity, as these are currently the most important factor that is likely to impact on prison population changes in the short/medium term. The model does not currently simulate flows for different crime-types, so crime-based trends are not explicitly modelled. As the model can be validated against historical data and can reflect a wide variety of dynamics, it is suitable for forecasting both in the short and medium term. However, there is uncertainty underlying the justice system's recovery (e.g., the rate at which court backlogs can be tackled) which impacts the model's assumptions and longer-term predictive power. Also, since the model works with individual data, as the projections extend further into the future there is greater uncertainty in the level of registrations of cases for prosecution in court. Due to these effects, currently only a 6-month projection is provided.

The rest of this report is organised as follows. Section 3 provides a brief overview of prison population trends from 2010 to 2023. Section 4 includes a contextual narrative of the main drivers of changes in the prison population in the short-term. It provides a brief summary of how the key drivers of trial courtroom capacity have changed recently, and the expected changes in the short/medium term. Section 5 includes an overview of why and how Scottish Government prison population projections have been revised to account for post Covid-19 pandemic justice system recovery. It provides an overview of the assumptions used to project the prison population and explains how the model's previous projections are validated for robustness against actual prison population figures in recent months. Section 6 summarises the population projection results for the period from April 2023 to September 2023. Section 7 concludes the report and briefly discusses other factors which may influence the size of the prison population in the longer term.

This report is the first of its kind and it is envisaged that it will be updated bi-annually (in Spring and Autumn). Each update will include an assessment of how the model has performed by comparing the previous population projections and the actual population level over recent months. Continuous development of the model and user engagement will be carried out to ensure appropriate and timely adjustments are applied to the model's assumptions and capabilities, in order to maximise its value in informing decision making and capacity planning of justice partners.