Criminal Proceedings in Scotland: 2023-2024

Spotlight: what has driven the long-term rise in average custodial sentence length?

The following section is designated as Official Statistics. The statistics here meet the standards set by the Office for Statistics Regulation in the Code of Practice for Statistics, but have not yet undergone the accreditation process.

Introduction

In this new section of the bulletin, we pick one area of criminal proceedings statistics to explore in greater depth. The topic chosen for this year is average custodial sentence length[1]. This choice was made for multiple reasons. It is a clear long-term trend which likely has multiple underlying causes but can be easily misunderstood as a simple increase in sentence harshness. In addition, this analysis may facilitate a better understanding of sentencing more broadly and inform work being undertaken by the Sentencing and Penal Policy Commission.

Criminal Proceedings in Scotland statistics have shown a consistent rise in average custodial sentence length[2] over the long-term. Across the 10-year span (2014-15 to 2023-24), average custodial sentence length rose by 37%, from 286 days to 393 days. Year-on-year rises were seen throughout this span with the pandemic-affected 2020-21 the only exception (Table 10c).

The question we hope to address here is: what has driven this long-term rise?

Multiple factors are considered when a convicted person is sentenced, and these may influence the final sentence given in complex and overlapping ways. These are laid out in more detail in the Scottish Sentencing Council Sentencing Factors.

Following a similar analysis conducted by the Ministry of Justice[3] we restrict our analysis here to three main potential driving factors:

• offence mix proceeded in courts
• custody rate for convicted persons
• sentence inflation

These three factors were selected as they are both the most likely to affect average custodial sentence length and can be relatively easily measured using the data we hold[4].

Offence mix

Offence mix is simply the relative proportion of different crime types prosecuted in court. A greater proportion of serious offences can drive an increase in average custodial sentence length as these typically result in longer sentences. Such changes may be influenced by both the prioritisation of certain types of court cases and broader trends in what crimes are being committed, recorded by police and prosecuted by COPFS.

A role for offence mix in driving average custodial sentence length is already suggested by the proportional rise in solemn convictions (Sheriff Solemn and High Court). This is associated with a proportional fall in Justice of the Peace court convictions (Chart 5). Justice of the Peace courts deal with the least serious offences and are restricted to a maximum custodial sentence of 60 days. Additionally, the two most serious crime groups (Non-sexual crimes of violence and Sexual crimes) both buck the overall trend of falling prosecutions. Whilst the number of people prosecuted across all crimes and offences fell by 41% between 2014-15 and 2023-24, Non-sexual crimes of violence only fell by 24% and Sexual crimes rose by 26% (Table 4a).

Custody rate

Custody rate refers to the proportion of people convicted for a specific crime that are sentenced to custody. This may drive an increase in average custodial sentence length in different ways. Both increasing custody rates for crimes that attract longer sentences and decreasing custody rates for those that attract shorter sentences may drive an increase in the average. This latter possibility means the rise in average custodial sentence length could be driven by more, and not less, lenient sentencing. The 2010 Presumption Against Short Sentences (PASS) and the 2019 extension of PASS specifically aimed to lower the number of short term sentences for less serious crimes.

A role for custody rate in driving an increase in average custodial sentence length is suggested by a proportional fall in shorter sentences. In 2014-15 sentences of 12 months or less made up 82% of all custodial sentences. By 2023-24 this proportion had fallen to 73% (Table 10d).

Sentence inflation

Sentence inflation is when increasingly longer sentences are given for the same crime. This can contribute to a rising average custodial sentence length and is the simplest interpretation of the overall rise. Sentence inflation is possible as for most crimes both legislation and sentencing guidelines permit judicial discretion. For example, aggravating factors may lead to a harsher sentence and mitigating factors a more lenient one. Some judges choose to publicly explain their sentences and interested readers may wish to consult the Judiciary of Scotland's Sentencing Statements for examples of these.

A role for sentence inflation in driving an increase in average custodial sentence length is suggested by trends amongst the most common crimes. These will have the greatest influence on the average. Some of the highest numbers of custodial sentences are for a main charge of Common assault, Crimes against public justice and Threatening and abusive behaviour (Table 9a). Between 2014-15 and 2023-24, average custodial sentence lengths for these offences rose by 17%, 20% and 19%, respectively (Table 10c).

Methodology

Criminal Proceedings in Scotland statistical bulletins only publish a 10-year span of data. If we were to choose this period for analysis, then pandemic effects would form 40% of the time series (2020-21 to 2023-24). We therefore chose to use a longer run, covering all comparable data from 1989-90 to 2023-24. Chart S1 shows the long-term rise in average custodial sentence length spans at least 34 years, from a minimum in 1990-91 to the 2023-24 maximum.

Chart S1. The 35-year trend for average custodial sentence length shows a long-term rise from less than 200 days to almost 400 days.

Average custodial sentence length in days (excluding life and indeterminate sentences), 1989-90 to 2023-24.

Analysis was conducted at the “crime code” level. There are around 600 crime codes used in the current statistical crime classification. These are aggregated into the 50 crime categories reported in the bulletin. For example, the category “Threatening to or disclosing intimate images” includes two distinct crime codes: “Threatening to disclose an intimate image” and “Disclosure of an intimate image”. We chose to use crime codes to keep as much detail on crimes as possible, and still have enough people in each group to be statistically useful.

For each crime code and financial year, we tabulated the following information:

• the number of convicted people
• the proportion of convicted people given a finite custodial sentence
• the average duration of any finite (i.e. excluding life and indeterminate) custodial sentences

For each crime code we also generated an average value across the entire time span to use as a reference year. This differs from the Ministry of Justice approach, where 2010 was used as a reference year. This decision was made to account for the fact that new crimes are introduced over time. For example, crimes under the Sexual Offences (Scotland) Act 2009, Domestic Abuse (Scotland) Act 2018 and others did not exist in 1989-90 but now form a substantial proportion of custodial sentences.

Average custodial sentence length for all crimes across the 35-year timespan was then calculated following these steps:

1. Take the average number of people convicted of a particular crime.
2. Multiply this by the average custody rate for that crime, giving the average number of people given a custodial sentence for that offence.
3. Multiply the value from step two by the average custodial sentence length for that crime, giving the total sentence length for that crime.
4. Sum the values from step three for all crime codes to give the total custodial sentence length across all crimes and offences.
5. Divide the value from step four by the sum of the values from step two to give an average custodial sentence length across the 35 years.

The resulting value (265 days) was then used as a “null” model against which to compare the performance of our three sole-driver models (Chart S2). These three models were each generated by substituting the average value in the above steps with (for each financial year):

• the actual number of people convicted of that crime (to test offence mix)
• the actual custody rate (to test custody rate)
• the actual average custodial sentence length (to test sentence inflation)

Chart S2. A visualisation of the methodology used here shows how the null, offence mix, custody rate and sentence inflation models differ.

A visualisation of the core equation used for each model.

Data limitations

Data from Criminal Proceedings in Scotland statistics does not include all the factors that can influence an individual sentencing decision. Some of these factors are simply not present in the data. For example, criminal proceedings statistics holds no information on:

• whether a person was convicted of further charges (which can lead to a longer sentence)
• whether a person pled guilty (which can lead to a discounted sentence)
• a person’s employment history, caring responsibilities or other mitigating factors (which may lead to a shorter sentence)
• a person’s culpability or how involved or responsible they were for any offending (which may lead to a longer sentence)

Other details are present in the data but are too complicated to easily include in the models presented here. These include:

• any aggravations against the main charge (which can lead to a longer sentence)
• a person’s previous convictions (which can lead to a longer sentence)

Aggravations are complicated as a single charge can carry none, one or multiple aggravations and these can be statutory or non-statutory (see the aggravations section). They can also have complex and counterintuitive relationships with sentence length. For example, a bail aggravation (indicating a breach of bail conditions) can lead to up to six months being added to a custodial sentence. However, a serious offence may lead to a person being remanded in custody rather than bailed. This means bail aggravations can be associated with shorter rather than longer sentences.

A person with previous convictions may receive a longer custodial sentence. How relevant any previous conviction(s) are to a sentence is not simple to predict. This is especially true if the previous offending was very different in nature. Criminal proceedings data is also used to calculate how often people are reconvicted. This is published separately as Reconviction Rates in Scotland statistics. This data does not include all previous convictions. For example, convictions prior to 1989-90 or offences committed in other countries (including other parts of the UK) are missing.

These limitations should be borne in mind when considering the results presented here.

Results: offence mix

A model in which offence mix is the sole driver of average custodial sentence length does a good job[5] of explaining the rise in average custodial sentence length (Chart S3). This model shows the same broad rising pattern seen in the actual data and does especially well in capturing some of the most recent fluctuations potentially attributable to pandemic shifts in case mix. However, the model also consistently overestimates average custodial sentence length in the early part of the time span (from 1989-90 to around 2008-09). It then underestimates average custodial sentence length in the later part (from 2010-11 to 2023-24).

Chart S3. A model where offence mix is the sole predictor of average custodial sentence length shows a broadly rising trend and captures the pandemic-related fluctuations of recent years.

Performance comparisons of the null (solid red line) and offence mix (dashed black line) models against observed average custodial sentence length (solid black line).

Results: custody rate

A model in which custody rate is the sole driver of the rise in average custodial sentence length is a poor predictor[6] (Chart S4). Not only does it predict a declining trend – the opposite of what we observe – it performs worse than our “null” model that simply uses the overall average across the 35-year span.

Chart S4. A model where custody rate is the sole predictor of average custodial sentence length shows a broadly declining trend.

Performance comparisons of the null (solid red line) and custody rate (dashed black line) models against observed average custodial sentence length (solid black line).

Results: sentence inflation

A model in which sentence inflation is the sole driver of average custodial sentence length does a good job[7] of explaining the rise in average custodial sentence length (Chart S5). This model also shows the same broad rising pattern seen in the observed data and does especially well in capturing some of the fluctuations in the middle portion of the time span, particularly the dip seen around 2006-07. However, the model does less well in the early and later parts of the time span. Average custodial sentence length is overestimated in the first few years and underestimated in the most recent (pandemic affected) years.

Chart S5. A model where sentence inflation is the sole predictor of average custodial sentence length shows a broadly rising trend but does a poor job of capturing the pandemic-related fluctuations of recent years.

Performance comparisons of the null (solid red line) and sentence inflation (dashed black line) models against observed average custodial sentence length (solid black line).

Conclusion

When considered as sole predictors of average custodial sentence length both offence mix and sentence inflation show increasing trends broadly in line with observed data. However, neither model makes a perfect prediction, and their strengths and weaknesses differ, suggesting both play an important role.

By contrast, custody rate is a poor sole predictor. Combined with a clearer role for sentence inflation this favours an overall trend of increasingly harsher, rather than increasingly lenient sentencing.

However, as a rising trend in average custodial sentence length can be explained by offence mix alone (i.e. when sentencing harshness is in effect held constant) it is not possible to infer that increasingly harsh sentencing explains the entirety of the observed rise. In addition, because the data presented here only considers the main charge[8], and has other limitations[9], we should be cautious in assuming sentence inflation is necessarily occurring. It could be that a rising number of charges for convicted persons or some other trend amongst sentencing factors that cannot be accounted for here are at work.

Overall, the analysis presented here suggests that the rise in average custodial sentence length is not driven purely by increasingly harsh sentencing, but nor is there a clear role for increasingly lenient sentencing.