By Troy Riddell, July 10, 2026
Bail has become one of the most contentious issues in Canadian criminal justice. Critics point to cases where individuals released on bail commit serious offences, while advocates warn against unnecessarily detaining legally innocent people.
In response to public pressure from premiers, police organizations, and opposition parties, the federal government recently passed Bill C-14 to tighten aspects of Canada’s bail regime.
Yet these debates are taking place absent robust, widely available, and consistent cross-country bail data. Governments do not systematically collect or publish basic information about who is granted bail, under what conditions, or whether those individuals comply with release orders or commit new offences while awaiting trial. As a result, policymakers, practitioners, and the public are often forced to rely on anecdotes, isolated tragedies, and competing political narratives rather than comprehensive data.
Consider a recent case in Kelowna. An accused charged with threatening to kill his former partner whom he had previously been convicted of assaulting recently was granted bail on strict conditions, including a no-contact order, electronic monitoring, residence at a treatment facility, and a prohibition on returning to Kelowna. The judge who approved the plan, which was consented to by the Crown prosecutor, noted ensuring that the accused received “treatment for his mental health and addiction issues is ultimately in the ‘long-term interest of public safety.’” Electronic monitoring and living in a different community, the judge continued, would help to address “real safety concerns” and provide some peace of mind to the complainant. The accused was still in remand five days after being granted bail presumably as arrangements of the bail plan were being finalized.
This one example illustrates the range of critical decisions that are taken at each stage of the bail process, including:
- Whether an individual should be released by police or sent to a bail hearing.
- Whether the Crown should consent or contest a bail release by “showing cause” why the accused should not be released (for certain kinds of offences the onus is reversed and the accused must show why they should be released).
- Whether a Justice of the Peace or judge should release an accused on bail – and if so, with what conditions?
For guidance, judges must consult the Criminal Code along with judicial precedents. They must also consider the constitutional right to reasonable bail found in the Charter of Rights and Freedoms. But few of these “judgement calls” are based on hard data.
The Carney Liberal government’s reform package in Bill C-14 includes:
- Adding more offences subject to “reverse onus.”
- Clarifying that the “principle of restraint” does not mandate release.
- Stipulating that the “ladder principle” does not apply to accused subject to reverse onus.
- Requiring courts to consider whether the allegations against the accused involve random or unprovoked violence.
The Conservative Justice critic argued that the changes would be ineffective as the bill does not repeal the principle of restraint, while defence lawyers, the Canadian Civil Liberties Association, and the John Howard Society, among others, questioned the constitutionality of the legislation; doubt it will increase public safety, and worry that it would perpetuate inequities in the bail system.
Unfortunately, these important debates about bail are taking place in an empirical vacuum. This dearth of data makes it difficult to reform our bail system to make it more efficient, effective, and just. It is important to examine the data we do have, where the gaps are in our knowledge, and how the limited data we have is often presented in different ways. This all points to the different paths forward, as well as potential improvements to data collection.
Police discretion to release an accused prior to a court appearance
The bail process begins when police decide to hold an accused for a bail hearing rather than releasing them prior to their court appearance on their charges. Criminologists have argued that police are increasingly prone to hold people for a bail hearing rather than releasing them until their case is resolved at trial (or through a plea agreement). According to Webster et al., between 2001 and 2007, the per cent of cases that began with a bail hearing in Ontario grew from 39.2 per cent in 2001 to 50.2 per cent in 2007. More recent statistics, however, suggest a little more than 40 per cent of cases being referred to bail court by the police — not indicative of a sustained increase in police detention. We do not know, though, if there are any patterns as to who is sent for a bail hearing — presumably the nature of the alleged crime and the accused’s criminal history will be factors, but are there other influences on police decision-making?
Bail hearing outcomes
Unfortunately, once individuals are held for bail, we lack good data on the results of bail decision-making and why those decisions were made. Not infrequently, commentators will point to rising numbers of individuals held in “remand” — awaiting the disposition of their case through a plea bargain or trial/sentencing — as a proxy for how many people are being denied bail. Using remand numbers in this way is problematic, however, as remand counts are influenced by a complex array of factors.
Today, accused individuals are spending more and more time in remand due to a range of variables, including the number of people being sent by police to bail hearings, multiple adjournments during the bail process, inefficient use of court time, and cases in which bail decisions are not rendered.
In the case of the Kelowna accused, the number of days the individual spent in remand before and after the decision to release will count towards remand — but his time should not contribute to an argument that bail is being denied frequently.
Another way of presenting the data is to highlight release rates, as a recent Globe and Mail article did when discussing the recent Liberal government reforms in October 2025. Using data collected by the Ontario Court of Justice (OCJ), the article noted that only 50 per cent of individuals were released in 2024 compared to 57 per cent in 2018 . People may assume that means almost half of all accused who have a bail hearing are denied bail. However, that is not the case. As the chart in the Globe and Mail article reveals, there is a large percentage of bail cases in Ontario with no outcome. Indeed, when the Globe and Mail examined the number of bail denials as a percentage of all bail cases in an April 2026 article – the rate of denial in 2025 was 4.7 per cent. (It is unclear if other provinces have a similarly large group of people that seem to be avoiding a bail decision – nor do we know why these cases exist. It could be that the accused has no permanent residence, or perhaps, they simply prefer to stay in remand, where time accumulated typically counts as 1.5 times credit for time served).
If one omits the large number of Ontario cases in which there is no decision in the OCJ, the rate of denial in cases in which there is a bail decision (as opposed to all bail cases as used in the Globe and Mail) ranged from 5.1 per cent (2018) to 8.3 per cent (2025).
The OCJ data represent a rare example of the systematic collection of bail statistics that are publicly available. Researchers have taken to directly observing bail hearings to better understand outcomes. The CCLA observed bail courts in British Columbia, Manitoba, Ontario, and Nova Scotia between November 2021 and November 2022 and noted that on an average day adjournments were the most frequent outcome (66 per cent); however, only 2 per cent of individuals were denied bail when a decision was made (there was variability between provinces).
The most recent CCLA report notes that the vast majority of individuals released on bail have conditions attached to their release, such as weapons prohibitions, not contacting the alleged victim, and residency requirements. The report notes that since the last report patterns of conditions have changed, such as Ontario requiring fewer sureties or treatment programming, though regional variations remain. Similarly, Yule and Schumann indicate that following the Supreme Court’s decision in Antic (2017), which reinforced the ladder system and principle of restraint in the Criminal Code, the number of conditions declined in three Ontario court locations.
What factors explain bail outcomes?
Data about what factors correlate with the denial of bail are even more difficult to find. The BC Prosecution Service had to manually scour files to collect bail data – its research suggests detention rates were slightly higher than average when there was only a violent offence involved (between 10–13 per cent) and notably higher where there was a combined violent offence and breach of conditions (between 17 and 24 per cent). A 2013 study prepared for Justice Canada found that unemployment and prior criminal history were statistically significant indicators of being detained in remand at three de-identified sites. The researchers found that offence type played a role in decision-making, but that seriousness of offence did not always correspond to a greater chance that an accused would be denied bail (p. 18). For instance, 53 per cent (n=30) of break and enter cases were denied bail, compared to 32.2 per cent (n=233) of assaults (all types except sexual assault) and 55.6 per cent (n=9) of fraud cases. Release rates have also been associated with demographic characteristics of the accused, with observed differences in release by gender and race even when factors like criminal history are taken into account.
Behind these statistical correlations are organizational factors that impact bail decision-making. There is provincial variation in bail processes; for example, some provinces use Justices of the Peace (JPs) almost exclusively, whereas other provinces have Justices of the Peace oversee proceedings where the Crown does not oppose release while provincial court judges decide cases where there is a contested bail hearing. Perhaps not surprisingly, defence counsel report that individual Crowns and JPs/judges can vary significantly in their willingness to release accused and under what conditions.
Tragic incidents have also called into question the quality of the files being presented to decision makers. The Irving Report (2016) in Alberta, for instance, prepared following the shooting of two police officers (one fatally) by an individual on bail, called for greater professionalization of bail processes, including the need for more complete information to be provided at bail hearings. Although the accused was not in Alberta’s Priority Prolific Offender Program (PPOP) (likely owing to budget constraints), Irving argued that a dossier later developed by the PPOP office on the accused contained information that would have been useful during the bail hearing. More recently, questions were raised about why individuals who committed horrific stabbings in the Toronto subway had been released with outstanding warrants against them.
It raises the question: Are these isolated incidents or do individuals released on bail commit serious offences?
Offending while on Bail
Regrettably, data is even more sparse on this key question. As the CCLA notes, federal and provincial governments do not keep records around offences committed by individuals on bail. StatsCan records administration of justice charges, which include violations of bail orders (as well as other violations such as peace bonds and parole), but we have no systematic data about individual offending while on bail. Winnipeg Mayor Scott Gillingham started producing a weekly bulletin describing crimes committed by individuals under court order, including bail. The first bulletin in September 2025 described an individual accused of a shooting who was released on bail and then allegedly engaged in a carjacking. The bulletin also offered statistics from August 2025 indicating that 176 people who were subject to bail or probation conditions committed offences, including seven assault charges against individuals who were not to be contacted. It is unclear, however, how those statistics were created. Both Manitoba Justice and the Winnipeg Police Service acknowledged to the CBC that they do not systematically track such statistics. In the same article, the Chief Judge of Manitoba’s provincial court argued that only a “tiny fraction” of 40,000 charges were committed by individuals on bail – but he could not point to statistical data to verify his claim.
We are mostly left to rely on media accounts of individuals violating their bail or ad hoc studies constrained by jurisdiction and time. Examples of the latter can be found in data presented to the Senate Standing Committee on Bill C-14 by the Commissioner of the Ontario Provincial Police (OPP), on behalf of the Canadian Association of Chiefs of Police. The submission notes inconsistencies across police organizations in how they collect information. The data from Toronto and York Regional police appear most closely tied specifically to bail (as opposed to possibly also incorporating other forms of release). The rate of reoffending in the two regions ranged from 33 per cent to 49 per cent per year (from 2023 to 2035). Failure to comply with court orders made up a large portion of charges while on bail, though there were new charges as well. For instance, in Toronto almost half of the individuals had a “failure to comply” violation, while another third of the individuals had a “failure to comply” along with a new substantive charge (and 22 per cent had a new charge only). On a more granular-level, Toronto Police Service report that 7 of 44 gun-related homicides were committed by individuals on bail in 2022.
The Toronto and York figures for recidivism are higher than some other studies. In a now deleted webpage, the Government of Alberta reported that “27.9 per cent of accused adults under bail supervision were admitted to remand custody at least once due to violating bail conditions and/or incurring new charges between 2021 and 2022”; however, no other contextual data were provided (for instance, there were no specifics on the types of new charges – were the individuals late for curfew, or accused of a violent home invasion?). The 2013 report for Justice Canada was only able to collect recidivism data from two of the four jurisdictions it studied and found that 17.5 per cent of individuals released on bail re-offended, but the vast majority were for breaches of conditions or not showing up for court, with only 2 per cent committing new substantive offences.
Perhaps the York and Toronto figures feature more offending while on bail than other studies because York’s data are for individuals with “street enforceable conditions” and Toronto’s are for accused charged with serious offences. Or maybe there has been an increase in serious offending while on bail. We simply don’t know due to the inconsistent methodologies involved in these ad hoc analyses.
In addition to not knowing how generalizable findings from these studies are, we also have no contextual data about individuals on bail (demographic and socio-economic data, conditions, social services they used, etc.). That means we do not know what factors are associated with either recidivism or successful bail.
Analysis
Some may argue that this lack of data is not that important — we could simply instruct judges/justices of the peace to be more cautious about allowing bail, especially for repeat, violent offences, as Bill C-14 aims to do. Although such a step may have benefits, there are limits to this approach. First, constitutionally, decision-makers are always going to have to balance the presumption of innocence and the right to reasonable bail against public safety and the reputation of the administration of justice. Second, it is possible that increased use of remand before trial could undermine public safety in the long run as individuals are separated from families, employment, social services — supports that encourage more pro-social behaviour. Third, housing individuals in prison before a resolution of their charges is expensive. Fourth, better data could also reveal inefficiencies, delays, and any patterns of unfair treatment in the bail process. Finally, having quality data throughout the process would help address debates about whether there is a culture of “risk aversion,” or whether the system is “catch and release.” The limited qualitative and quantitative data we do have suggest both narratives offer an over-simplified view of Canada’s bail system.
The following types of data could aid legislators and front-line decision-makers better protect public safety while upholding fundamental principles such as the presumption of innocence. It would also help identify inefficiencies and unfair treatment and support more effective allocation of policing, social services, correctional resources:
- Nature of the charges (how many and what specific Criminal Code or other provisions).
- Individual and social characteristics of each accused (gender, race, employment status, housing status, marital status, number of children, caregiver status).
- Criminal involvement of the accused (criminal history, outstanding warrants).
- Nature of the bail hearing(s) (length of time, whether the Crown consented to release, whether a reverse applied).
- Bail decision (bail denied or release granted; if granted, types of conditions and whether service providers are involved).
- Post-bail behaviour (whether the individual reoffended while on bail; if so, what were the nature of the charges? Here it would be helpful to disaggregate “failure to comply” charges into ones that are more innocuous (e.g. breaking curfew) versus ones that can have a serious impact on victims (e.g. making contact with a complainant).
To the extent that AI may be introduced to aid decision-making in criminal justice, it is particularly important to have quality data. Regardless of who — or what — is involved in the decision process around bail, more precise data could aid the current system in being more efficient, effective, and just. Ultimately, even with better data there will still be normative disagreements amongst reasonable people about whether bail decisions should err on the side of caution or presumption of innocence/liberty of the accused – currently, though, the intertwined empirical and normative debates are occurring within a vacuum of evidence.
The brief of the Canadian Association of Chiefs of Police to the Senate Standing Committee on Legal and Constitutional Affairs argued that “[d]ata availability” is key to understanding the implications of Bill C-14. The submission highlighted that “[f]undamentally, national data remains limited in terms of collection, availability and standardization. An alignment of data from law enforcement, courts, and corrections is required to fully address key aspects of bail, reoffending, and public safety.” Bill C-14 requires the federal Minister of Justice to annually present data on: “judicial interim release outcomes, including in respect of compliance with release conditions, recidivism by accused at large on release orders and incidents requiring public attention; b) an analysis of the effectiveness of release conditions; and c) data on the accessibility of judicial interim release and disparities between different groups.”
A federal-provincial/territorial committee is working to help produce better data, but it will face considerable obstacles. As discussed above, systematic data collection is relatively rare. Moreover, within each province there are different organizations that are or could collect data, but standardization and sharing is uncommon between ministries of Justice/ Attorneys General, ministries of the Solicitors General/Public Safety, the courts, and various police agencies.
The fractured nature of the criminal justice “system” also contributes to poor data about all facets of criminal justice in Canada from policing to sentencing to corrections. One consequence is that one cannot trace individual outcomes through the system. Are individuals who violate bail conditions more likely to reoffend while later serving a sentence in the community? Should there be more local police focus on such individuals? How many and what kinds of individuals are held on bail but do not reoffend while serving part of their sentence in the community — would it have been fairer and more cost effective to have released (some) of these individuals at the bail stage?
Although the “administration of justice” falls under provincial jurisdiction, it may be useful for the federal government to offer funding to provinces to help with upgrading and standardizing data collection in criminal justice. Even in the unlikely event that there were funding and all jurisdictions agreed to cooperate, however, the project would be a long-term one involving discussions around the data to collect, hardware and software, data ownership, and privacy/ data access issues.
In the meantime, more modest but meaningful data collection efforts should be undertaken. For instance, police forces can build upon their nascent data collection efforts to trace all or a sample of accused who are charged: types of charges, demographic characteristics, whether an accused was on bail at the time (or some other form of release), and whether the accused was released on their own recognizance or sent for a bail hearing. Other provincial judicial systems could follow Ontario’s lead in publishing provincial court statistics about bail. Justice Canada could conduct a follow-up study to its useful but dated 2013 report that includes statistical analyses of release and denials and recidivism while on bail — hopefully more jurisdictions could be included in the future. Groups who operate bail bed supervision programs could undertake program evaluations that include recidivism data. These smaller data collection efforts could not only provide us with much needed context about bail, but they could also provide guidance for longer-term data collection efforts.
The current lack of data around an issue that has profound implications for justice, community safety, victims, and the rights of accused is deeply troubling and needs to be addressed.
Troy Riddell is associate professor and chair of the Department of Political Science at the University of Guelph. He has interests in judicial politics and criminal justice policy and is a co-author of Canadian Courts: Law, Politics, and Process.
