Skip to main content
SearchLoginLogin or Signup

Preregistration of Study into Disparities in Remand Decisions in England and Wales

Published onMay 30, 2024
Preregistration of Study into Disparities in Remand Decisions in England and Wales


This is a preregistration of a study that will investigate the degree of (in)consistency in remand decisions between police forces and between courts in England and Wales. It will do so using new criminal court datasets made available by the Data First project from the Ministry of Justice (MoJ) in collaboration with the Office for National Statistics (ONS) and Her Majesty Courts and Tribunal Service (HMCTS). The study will also use these linked datasets to assess how disparities in remand rates evolve throughout the trajectory of cases from the police remand decision up to the final Crown Court decision, when cases makes it that far.

Literature Review and Hypotheses

When a person is charged with a criminal offence, a decision must be made as to whether he/she will be remanded in custody or released on bail (with or without conditions) pending trial. A person’s bail/remand status will be determined by multiple decision makers over the duration of the legal proceedings. First by a police custody officer, then by a magistrates’ court bench or a District Court Judge and then, if the charge is being dealt with on indictment, by a Crown Court Judge. Bail will also be considered whenever there is a material change in the circumstances of the defendant and on any instance the case comes before a court. The PACE Act 1985 (governs police bail decisions) and the Bail Act 1976 (governs court bail decisions) both contain a presumption in favour of unconditional bail. However, bail may be refused by the police or by a court if there are ‘substantial grounds’ for believing that the person would - if released on bail – fail to appear in court/surrender to custody, commit an offence while on bail, interfere with witnesses or otherwise obstruct the course of justice. Bail may also be refused if there are reasonable grounds for believing it is necessary for the person’s own protection (Section 38(1) of PACE Act 1985; Schedule 1 of the Bail Act 1976).

The importance of the decision to grant bail or remand in custody cannot be overstated. First and foremost, to refuse bail and remand in custody, is to deprive a person who has yet to be convicted of the offence charged, of their fundamental right to liberty. At a personal level, being remanded in custody can trigger a wide range of negative consequences. For example, those held in remand are at a risk of losing their accommodation and their employment as well as their support networks (Justice 2023). This not only impacts the accused person but may also have profound implications for the person’s family and dependants. This is made particularly striking if we consider that 10% of those remanded in custody by a court are subsequently acquitted of the offence charged (House of Commons Justice Committee 2023, 29). To make it worse, those who are acquitted and released following a period of remand do not get access to community-based supports upon release nor do they receive any financial compensation.

In addition to the 10% acquitted, a further cohort - 63% of those remanded and tried in the magistrates’ court in 2020 - do not end up receiving an immediate custodial sentence for the offence of which they are convicted (Transform Justice 2022). For some, this will be due to time spent remanded in custody being equal or close to the appropriate sentence for the offence committed, so a further immediate custodial sentence is not warranted as it has in effect already been served. However, that 63% will inevitably also include individuals who spent longer on remand than the appropriate sentence of imprisonment warranted, as well as those whose offence does not warrant a custodial sentence at all. It is likely that this particular problem has grown in recent years. As backlogs in the criminal justice system steadily increase, so too does the time individuals spend on remand. As of September 2022, almost a third of the remand population had been held beyond the 6-month custody time limit (House of Commons Justice Committee 2023, 15). This figure has gone up from about 11% a decade earlier (HM Inspectorate of Prisons, 2012, 34).

Furthermore, besides its direct personal impact, there is evidence that being remanded in custody can have knock on negative effects at later stages in the criminal justice process. Being held on remand can impact a person’s right to a fair trial for the offence charged, due to practical obstacles in accessing a lawyer and facilities to prepare a defence (Smith 2021; HM Inspectorate of Prisons 2009). For those convicted of the offence charged, it may increase the severity of the sentence imposed, with those who have been remanded in custody being 7.5 times more likely to receive a custodial sentence than those who were granted bail (Lymperopoulou 2022).

Yet, despite its importance, there has been a distinct lack of research attention on the subject of remand over the years (Smith 2021), which becomes particularly stark when compared to other, more visible, stages of the criminal justice process like sentencing. Consequently, there remains significant gaps in our understanding of remand decision making. This is particularly so in England and Wales, where only a handful of empirical studies have been undertaken. Findings from these studies show that there are numerous issues with the legal framework and operation of remand procedures that raise questions about the rigor and consistency of remand decisions.

Dhami (2010; 2005) and Smith (2021) highlighted the vagueness of the laws governing remand decisions and the considerable discretion this affords to both magistrates and judges in determining whether to remand a person in custody. Magistrates and judges have discretion in relation to the information that can be considered when assessing the risk of absconding, offending and obstructing justice. They also have discretion as to how the information considered ‘relevant’ should be weighted as well as regarding the degree of ‘risk’ required to establish ‘substantial grounds’ for denying bail. The vagueness of the laws and the high judicial discretion can lead to inconsistencies in how the law is applied. For example, Dhami (2010) reported significant variability in magistrates’ interpretation of ‘substantial grounds’ and the degree of risk thought necessary to meet this criterion.

Secondly, concerns have been raised that remand hearings are often too brief and insufficient time is taken to fully consider each decision (Cape and Smith 2016; Justice 2023; Dhami and Ayton 2001). Cape and Smith’s (2016) court observations found that the average time of a prosecution submission was 2.5 minutes and 2 minutes for defence lawyers, raising to 5.6 minutes for defence lawyers in cases where remand was requested by the prosecution. Justice (2023, 13) highlighted similar issues with the length of time spent on remand hearings, noting in particular two cases observed by researchers where a decision to remand a person in custody took just 2 minutes. It was also noted that in 17% of cases where an individual was remanded in custody, the hearings lasted less than 10 minutes.

Thirdly, it has been shown that magistrates and judges are often required to make remand decisions with limited information, usually relying on the information provided by the prosecution during the remand hearing (Smith 2021). Bail Information Schemes were established to provide verified information on what bail conditions could be imposed and what support is in place to ensure adherence to these conditions. This information is vital in allowing defence lawyers to advocate for bail, and for magistrates and judges to make informed decisions. However, due to different problems in accessing Bail Information Schemes, this information is often not available to defence lawyers and decision makers (House of Commons Justice Committee 2023; Cape and Smith 2016). A survey by Transform Justice (2018) found that 37% of lawyers surveyed had never been contacted by bail information officers and 46% said they were ‘very rarely/not often’ contacted by them.

In addition to the above findings which, considered together, raise concerns regarding consistency, a few other empirical studies have specifically explored the issue of between courts consistency in remand decisions in England and Wales. They are considerably outdated, although they all document a similar finding of substantial between court variability in custodial remand rates. Jones (1985) analysed official statistics and found the proportion of defendants remanded in custody by magistrates’ courts varied from less than 10% (Bedfordshire 7%, Gwent 8%) to over 30% (Dorset 31%, North Yorkshire 37%). Hucklesby (1997) reviewed a sample of 2,069 court registers detailing magistrates’ remand decisions in three South Wales magistrates’ courts. She found two of them had custodial remand rates of 9% whilst the third had a rate of 25%. Morgan and Henderson (1998) analysed the remand rates for the five areas and found custody rates ranging from 16% to 21%.

Without an adequate level of consistency - i.e. if like cases are treated differently without justifiable reasoning - the fundamental principle of equality under the law is undermined, and with that a just legal system (Steyn 1997). Consistency is important in all legal decisions but it is particularly so in decisions to remand a person in custody while still presumed innocent, as this is one of the most consequential legal decisions.

In spite of the documentation of disparities in remand rates between courts in the literature, these findings have largely been disregarded by policy makers, likely in part due to methodological limitations. Jones (1985) did not control for important factors such as offence type and offender characteristics leaving open the possibility that disparities in remand rates could be the result of differences in the types of cases being dealt with in each court. Similarly, both Hucklesby (1997) and Morgan and Henderson (1998) had relatively small sample sizes with a limited ability to accurately control for a wide range of factors known to impact remand decision. Perhaps more importantly, all the empirical studies exploring disparities in remand in England and Wales are now over 25 years old, and the data used would have been collected prior to the Bail Act 1976 being amended by the Criminal Justice and Public Order Act 1994 and the Legal Aid, Sentencing and Punishment of Offenders Act 2012.

Another noteworthy limitation of the existing literature is that most studies have tended to focus on one particular stage of the remand decision making process, usually the initial magistrates’ court decision. As a result, we continue to have very little insight into how each remand decision impacts on the next and how remand rates evolve over the duration of a person’s case, from when first charged with an offence to the conclusion of proceedings. One partial exception to this is a study by Morgan and Henderson (1998), which considered levels of agreement between police decisions, prosecution recommendations and magistrates court decisions. It found significant agreement in cases where unconditional bail was granted but less so in cases where a person had been remanded in custody by the police. Importantly, however, this study was conducted prior to the amendment to Section 3 of the Bail Act giving police the power to grant conditional bail.

Our proposed study will address a number of existing gaps in the literature. We will use new criminal court datasets made available by the Data First project from the Ministry of Justice (MoJ) in collaboration with the Office for National Statistics (ONS) and Her Majesty Courts and Tribunal Service (HMCTS), to provide up to date and more valid estimates of unwarranted disparities in remand decisions. These are case-level administrative datasets capturing all hearings that took place at the Magistrates’ and the Crown Court in England and Wales from 2011 (2013 for the Crown Court data) to 2022. They contain 13,357,982 cases from the magistrates’ court and 1,000,827 from the Crown Court, and recording factors such as offence type, defendant characteristics (age, gender, ethnicity), remand status and court location. These data will allow us to control for a range of relevant factors and conduct a more accurate and reliable quantitative investigation into the degree of (in)consistency in remand rates between police forces and between courts in England and Wales than has previously been possible. Crucially, thanks to the linkage process undertaken by the Data First team, we will also be able to assess how disparities in remand rates evolve throughout the trajectory of the case from the police remand decision up to the final Crown Court decision, when the case makes it that far.

Leveraging the opportunities afforded by this new data, we will test the following hypotheses:

H1 - There are substantial disparities in remand rates between, police forces, magistrates’ courts and Crown Court locations in England and Wales, after controlling for case characteristics.

H2 - Disparities in remand rates will be more pronounced between police forces than between magistrates’ Courts, and between magistrates’ courts than between Crown Court locations.

Results ensuing the testing of the above hypotheses would have significant policy implications. If as we expect H1 is corroborated, we would provide evidence of inconsistencies that could not be entirely explained away by differences in the types of cases between courts, or dismissed on the basis that they are not representative of the decision-making process across the country. This would then contribute to move policy discussions forward towards considering and developing solutions to a problem that has long been thought to exist, but for which the evidence has never appeared to be sufficiently robust.

For example, one concrete policy solution we propose is the creation of a bail/remand guideline similar to those used in sentencing. The idea of introducing additional remand guidance for decision makers was recently suggested by the House of Commons Justice Committee (2023) but rejected by the Government in their response to the committee (2023), on the basis that it is not needed, which highlights how previous studies showing disparities in custodial remand rates currently hold little weight amongst policy makers.

There are many similarities between remand and sentencing decisions. Both involve subjective assessments and the weighing up of a wide range of factors. Both also require a balance allowing for individualised decisions while ensuring consistency between decision makers. In an effort to enhance consistency in sentencing in England and Wales, guidelines were introduced, first in the form of guidance and then in the form of definitive guidelines (Ashworth and Kelly, 2021). Recent empirical research has shown how sentencing guidelines in England and Wales have generally had a positive impact on consistency in sentencing practice (Isaac et al 2021; Pina-Sánchez and Linacre 2014; Pina-Sánchez and Linacre 2013) while also maintaining a degree of judicial discretion to allow the personal circumstances of the defendant to be taken into account and an individualised sentence imposed (Poppleton et al 2021; Roberts et al 2018). It would follow that a remand guideline, if properly constructed, could achieve the same goal.

If H2 is corroborated, it would highlight a clear and urgent need for research on police remand decisions in order to identify and understand the factors that are leading to disparities in remand decisions between police forces and weather these differ from factors impacting disparities in remand decisions between courts. This would be necessary in order to access if a remand guideline would be a suitable solution to disparities in police remand decisions as well as court remand decisions. While it could be argued that police remand is less consequential and in less need of attention on the basis that a person remanded in custody by the police will be brought before a court at the first possible opportunity for a hearing, disparities in police remand decisions could be contributing to disparities in court remand decisions. This could occur due to the court’s reliance on information provided by the police/prosecution and the deference of magistrates and judges to the prosecution recommendations (Justice 2023; Cape and Smith 2016; Morgan and Henderson 1998). The extent of agreement between police, magistrates’ court and Crown Court remand decisions will be assessed as part of our exploratory analysis.

Data and Modelling Strategy

The proposed study will be possible thanks to the new sentencing datasets made available by the Data First program. Data First is a research project funded by Administrative Data Research UK, linking datasets from across the justice system and other government departments, and making them available to accredited researchers via secure platforms1. Specifically, we will use the linked version of the first two datasets released by Data First, the Magistrates’ and Crown Court datasets. The former is sourced from extracts of Libra, the latter from XHIBIT, the administrative databases used by the Magistrates’ and Crown Court to manage cases across England and Wales (Jackson et al., 2022).

Our key outcome variable is remand. This is operationalised as a binary variable capturing whether the defendant is placed on remand as opposed to being granted bail (conditional or unconditional), and it is measured at two to three sequential stages, depending on whether the case is dealt entirely within the magistrates or committed to the Crown Court. For cases dealt within the magistrates’ court we will use: i) remand status set by police (‘remand_p_status_desc’ in the Magistrates’ Court dataset); and ii) remand status after the first hearing in the magistrates (‘remand_a_status_desc’ in the Magistrates’ Court dataset)2. For cases that end up being sentenced in the Crown Court we will use: i) remand status set by police (‘remand_p_status_desc’ in the Magistrates’ Court dataset); ii) remand status set by the magistrates court (‘bail_status_before_first_hearing’ in the Crown Court dataset); and iii) the remand status set by Crown Court after the last hearing (‘bail_status_after_last_hearing’ in the Crown Court dataset).

As part of our exploratory analysis, we will report changes in status of remand across each stage. We do not take this as a constitutive hypothesis of our study, given our focus on disparities, however, we believe that remand decisions may remain relatively stable across the system, with less than 10% of cases changing their remand status from the police decision to the Crown Court. We think this might be the case due to the deference of courts to prosecution recommendations and the high level of concordance between police, prosecution and courts previously identified (Hucklesby 1997).

To approximate ‘like with like’ comparisons in our analysis we adopt two approaches, we stratify our sample into specific offences, and we use statistical controls. We consider three specific offences, one for each of the main offence groups where considerations of remand are particularly relevant, i.e. offences against the person, drugs offences, property offences. The selected offences for each of those offence groups meet two criteria, they can be tried ‘either way’ – that is either summarily in the magistrates' courts or on indictment in the Crown Court - and they are processed in high volumes. These are: i) Assault Occasioning Actual Bodily Harm (Section 47 Offences Against the Person Act 186) totalling approximately 9,000 cases a year; ii) Supplying or offering to supply a controlled drug/ Possession of a controlled drug with intent to supply it to another (Misuse of Drugs Act 1971 s.4(3), Misuse of Drugs Act 1971, s.5(3)), roughly 18,000 cases a year; iii) Burglary (Section 9 of the Theft Act 1968), about 10,000 cases a year.

We will estimate a total of fifteen models, one for each combination of offences listed above with each of the remand variables considered for the analysis of cases processed in the magistrates and in the Crown Court. In all our models remand will be specified using a random intercepts binary logit approach, controlling for the same set of independent variables: offender’s age and sex, offence type, whether a guilty plea was introduced, and number of previous convictions recorded in the dataset. Age is a continuous variable, it will be centered around the mean and introduced as polynomial term of order two to capture the quadratic relationship between age and sentence severity reported in the literature (Ronald and Jacobs, 2002; Steffensmeier et al., 1995). For sex, we will use male as the reference category, and set unknown values as missing. After offence type, the number of previous convictions is likely the most consequential case characteristic in deciding remand or bail. Unfortunately, previous convictions is not directly recorded in the dataset. Instead, we will derive it for each case from the number of times an offender appeared before the hearing under consideration, in either the Magistrates’ or the Crown Court datasets, while sentenced to a disposal type different from an ‘absolute discharge’. To be able to follow offenders from the magistrates’ to the Crown Court we will use the ‘linked datasets’, the version of the sentencing datasets that provides a common unique offender identifier. We will be able to retrace previous convictions from as far back as 2011.

The random intercepts part of our models will capture the variability in the log-odds of remand across different police forces, magistrate’s courts, and Crown Court locations. To test H1 (There are substantial disparities in remand rates between police forces, between magistrates’ courts and between Crown Courts in England and Wales, after controlling for case characteristics) we will estimate the intra-cluster coefficient. Specifically, we will approximate the variance of level-1 residuals (referring to sentences) as π23\frac{\pi^{2}}{3} (Snijders & Bosker, 2011). We will declare H1 to be corroborated if the intra-cluster correlation coefficient is higher than 10%. This would represent that more than a tenth of the unexplained disparities in decisions of remand after controlling for relevant case characteristics stem from differences between police forces, magistrate’s courts, or Crown Court locations. To put that cut off point in context, the intra-cluster correlation coefficients reported in Pina-Sánchez and Linacre’s (2013) analysis of custodial sentence length decisions in the Crown Court ranged from 2.2% to 4.8% depending on the offence type considered. To test H2 (Disparities in remand rates will be more pronounced between police forces than between magistrates’ Courts, and between magistrates’ courts than between Crown Courts) we will use F-tests. Specifically, these tests will be used to determine whether the variance of the level-2 residuals (referring to police forces, magistrate’s courts, and Crown Court locations) is statistically significantly different when comparing police forces against magistrates’ courts and magistrates’ courts against Crown Court locations.

Further exploratory work will consider the effect of ethnicity on remand for each of the three stages considered. There is no legal justification for the defendant’s or offender’s ethnicity to influence ethnicity, hence why this variable is not introduced as part of the set of controls used to disentangle warranted from unwarranted disparities in remand. Instead, we explore the potential effect of ethnicity as a second stage part of our analytical strategy, where we move on from testing the relative presence of unwarranted disparities in remand to consider potential explanations behind those disparities. For this part of the analysis we will use defendant’s/offender’s ethnicity as ascribed by the police (‘ethnicity_police_defined_group’). We chose this particular measure over self-reported measures of ethnicity available in the dataset (‘ethnicity_self_defined_group’) since we want to explore potential discriminatory treatment, which would stem from the perceptions of criminal justice practitioners on the defendant’s/offender’s ethnicity (Pina-Sánchez et al., 2023). Just like it is for gender, the measure of ethnicity that we will use is affected by missing data to some degree. To adjust for this problem we will use multiple imputation. Specifically, we will use the MICE package in R (Van Buuren, 2018), to estimate five sets of imputations under Bayesian logistic regression, using the function ‘logreg’, and all the variables listed in this section (except for the three remand variables considered) as auxiliary data, plus another variable capturing self-reported ethnicity.


Ashworth, A. and Kelly, R. (2021), Sentencing and Criminal Justice, London: Bloomsbury Publishing

Bail Act, 1976

Cape, Ed and Smith, T. (2016) The practice of pre-trial detention in England and Wales: Research report, University of the West of England, Bristol. Available at: (Accessed: 19 February 2024)

Criminal Justice and Public Order Act, 1994

Dhami, M. K. (2010) ‘Lay magistrates’ interpretations of “substantial grounds” for denying bail’, The Howard Journal of Criminal Justice, 49, pp. 349-360

Dhami, M.K. (2005) ‘From discretion to disagreement: explaining disparities in judges' pretrial decisions’, Behavioral Sciences and the Law, 22, pp. 1–20.

Dhami, M. K., and Ayton, P. (2001) ‘Bailing and jailing the fast and frugal way’. Journal of Behavioral Decision Making, 14, pp. 141–168

Fair Trials (2016) A Measure of Last Resort? The practice of pre-trial detention decision making in the EU, London: Fair Trials, Available at: (Accessed: 19 February 2024)

HM Inspectorate of Prisons (2009) Remand Prisoners: A Thematic Review, London: HM Inspectorate of Prisons

House of Commons Justice Committee (2023) The role of adult custodial remand in the criminal justice system, Seventh Report of Session 2022–23, Available at: (Accessed: 19 February 2024)

House of Commons Justice Committee (2023) The role of adult custodial remand in the criminal justice system: Government Response to the Committee’s Seventh Report, Tenth Special Report of Session 2022–23, Available at: (Accessed: 19 February 2024)

HM Inspectorate of Prisons (2012) Thematic report by HM Inspectorate of Prisons: Remand Prisoners, Available at: (Accessed: 28 February 2024)

Hucklesby, A. (1997) ‘Court Culture: An Explanation of Variations in the Use of Bail by Magistrates’ Courts’, The Howard Journal, 36(2) pp. 129-145

Isaac, A., Pina-Sánchez, J. and Varela Montane, A. (2021) The impact of three guidelines on consistency of sentencing, Sentencing Council, Available at: (Accessed: 19 February 2024)

Jackson, T., Greyson, C., Rickard, I., Tseloni, A. (2022) Data First: Criminal courts linked data. an exploratory analysis of returning defendants and the potential of linked criminal courts data from 2011 to 2019 in England and Wales, London: Ministry of Justice

Jones, P. (1985) ‘Remand decisions at magistrates’ courts’, In D. Moxon (Ed.), Managing criminal justice: A collection of papers, London: HMSO, pp. 106–117.

Justice (2023) Remand Decision-Making in the Magistrates’ Court: A Research Report, London: Justice, Available at: (Accessed: 19 February 2024)

Legal Aid, Sentencing and Punishment of Offenders Act, 2012

Lymperopoulou, K (2022) Ethnic Inequalities in Sentencing in the Crown Court - Evidence from the MoJ Data First Criminal Justice datasets, ADRUK, Available at: (Accessed: 19 February 2024)

Ministry of Justice (2023) Criminal Justice Statistics quarterly, England and Wales, year ending December 2022 (annual). Available at: (Accessed: 18 March 2024)

Morgan, P. M., & Henderson, P. F. (1998) Remand decisions and offending on bail: Evaluation of the bail process project. HORS No. 184. London: Home Office

Police and Criminal Evidence Act, 1984

Pina-Sánchez, J. and Linacre, R. (2013) ‘Sentence consistency in England and Wales: Evidence from the Crown Court Sentencing Survey’, British Journal of Criminology, 53, pp. 1118-1138

Pina-Sánchez, J. and Linacre, R. (2014) ‘Enhancing consistency in sentencing: Exploring the effects of guidelines in England and Wales’, Journal of Quantitative Criminology, 30, pp.731-748

Pina-Sánchez, J., Geneletti, S., Veiga, A., Morales, A., & Guilfoyle, E. (2024) Can ethnic disparities in sentencing be taken as evidence of judicial discrimination? Journal of Legal Research Methodology, 3(1), pp. 54-82

Poppleton, S., Wedlock, E., Isaac, A. and Marshall, E. (2021) A Review of Consistency in Sentencing, Sentencing Council, Available at: (Accessed: 19 February 2024)

Roberts, J., Pina-Sánchez, J. and Marder, I. (2018) ‘Individualisation in sentencing: the effects of guidelines and preferred numbers’, Criminal Law Review, (2), pp. 123-136

Ronald, H. and Jacobs, D. (2002) ‘The political context of sentencing: An analysis of community and individual determinants’, Social Forces, 81(2), pp. 577–604

Smith, T. (2021) ‘‘Rushing remand’? Pretrial detention and bail decision making in England and Wales’, The Howard Journal, 60(1), pp. 46-74

Smith, T. (2022) ‘The Practice of Pre-trial Detention in England & Wales - Changing Law and Changing Culture’, European Journal on Criminal Policy and Research, 48, pp. 435-449

Steffensmeier, D., Kramer, J. and Ulmer, J. (1995) ‘Age differences in sentencing’, Justice Quarterly 12(3), pp. 583– 602

Steyn, K. (1997) ‘Consistency – A Principle of Public Law?’ Judicial Review, 2(1), pp. 22-26

Transform Justice (2022) The role of adult custodial remand in the criminal justice system: a response to the Justice Committee of the House of Commons, London: Transform Justice, Available at: (Accessed: 19 February 2024)

Transform Justice (2018) Presumed innocent but behind bars – is remand overused in England and Wales, Transform Justice, Available at: (Accessed: 19 February 2024)

Van Buuren S (2018) Flexible imputation of missing data, Boca Ratton: CRC press

No comments here
Why not start the discussion?