Reform, Then Reversal

Chart 1Nearly one in five people held on remand is Aboriginal

In 2025–26, 18.9% of adults received into remand in Victoria were Aboriginal, up from 13.2% in 2018–19. Aboriginal people make up about 0.9% of Victorian adults. The share kept climbing during the reform year: the reform cut non-Indigenous remand faster than Aboriginal remand. For women the share is higher still, at 25.6%.

Use the buttons to switch between all adults, women and men. Hover over a bar for the counts behind it.

Source: Crime Statistics Agency (2026), Corrections Victoria bail data tables, Table 01; population share from ABS (2024, 2026). A remand reception is one admission to prison on remand; one person can be received more than once a year.

Chart 2Per head of population, the gap is in the tens

Raw counts hide the size of the gap. Divide remand receptions by population and an Aboriginal adult in Victoria was about 26 times as likely as a non-Indigenous adult to be received into remand in 2025–26. That ratio was 18 in 2018–19, before the reforms. It is widest for women aged 35 and over, and it grew fastest for young women, from 9 times in 2023–24 to 32 times two years later. The numbers of young Aboriginal women are small, so single years swing widely, but the direction is clear.

Each cell is one group in one year. Switch between the rate ratio and the two underlying rates. Hover over a cell for the full numbers.

Sources: Crime Statistics Agency (2026), Corrections Victoria Table 01; Australian Bureau of Statistics (2024, 2026) population estimates and projections (Series B). Rates use the population at June of each year. People aged 18–19 are estimated as 40% of the ABS 15–19 group. Non-Indigenous population = total minus Aboriginal population.

Chart 3The switch: bail offences out, then back in

Why did the numbers turn so fast? One reason is mechanical. In 2024 the offence of committing an indictable offence while on bail was removed, and in March 2025 it was put back. Remand receptions involving a breach-of-bail offence fell from 3,389 in 2022–23 to 1,269 in 2024–25, then climbed back to 2,324 the year the offence returned. At the police station, bail decisions involving a breach charge almost doubled, from 5,886 to 10,770.

Switch between counts and shares. Hover over a bar for the numbers behind it.

Source: Crime Statistics Agency (2026), Victoria Police bail data tables, Table 04, and Corrections Victoria bail data tables, Table 03. Breach-of-bail offences include failing to answer bail, contravening a bail condition and committing an indictable offence while on bail.

Chart 4In court, young Aboriginal adults felt the change most

The CSA only began publishing court bail outcomes by Indigenous status in 2024–25, so the court record covers just the year either side of March 2025. Across the Magistrates’ Court the share of bail applications refused rose for almost every group. The biggest jump was for Aboriginal adults aged 18–24: from 28.7% to 41.4%, up 12.7 points. For non-Indigenous people of the same age it barely moved (32.6% to 33.5%). The Children’s Court runs the other way, a reminder that the pattern is not uniform.

Each line runs from 2024–25 (open circle) to 2025–26 (filled circle). Use the buttons to compare women and men.

Source: Crime Statistics Agency (2026), Magistrates’ Court and Children’s Court bail data tables, Table 02. Revoked bail is excluded from the refusal share. Children’s Court figures cover all ages appearing in that court, including a small number aged 18 and over.

Chart 5Back to where it began

Set the reform year as the baseline and only one group stands out. Since 2023–24, remand receptions of Aboriginal women have risen by 86%, from 120 to 223. Non-Indigenous women and men are roughly where they were. Switch the baseline to 2018–19, the year before Veronica Nelson died, and the reform’s uneven reach shows up: non-Indigenous remand is still well below its old level, while Aboriginal remand is almost back to it.

Change the baseline year with the buttons. Hover over a point for the count behind it.

Source: Crime Statistics Agency (2026), Corrections Victoria bail data tables, Table 01. Index = receptions in a year ÷ receptions in the baseline year × 100.

What it adds up to

Put the five charts together and one pattern repeats. The 2024 reform reduced remand for everyone, but least for Aboriginal people, the group the coroner’s findings were about. When the law tightened again in March 2025, Aboriginal remand rose first and fastest. An Aboriginal adult in Victoria is now about 26 times as likely as a non-Indigenous adult to be received into remand, up from 18 times before the reforms. Remand receptions of Aboriginal women have almost doubled in two years.

None of this shows whether tougher bail made Victoria safer; that question needs other data. It does show that bail law is not neutral in who it holds. Every bail promise made before 28 November can be checked against the same public tables used here, and the question to ask of each one is the one this report began with: when the law moves, whose numbers move with it?

What this does and does not show

These are counts of events, not people. One person can be remanded more than once in a year. Remand numbers reflect decisions made by police, courts and the law as much as anyone’s behaviour, so a higher rate is a measure of how the system treats a group, not a measure of the group. The data cannot show why individual decisions were made, and annual figures (July to June) blur the exact timing of each law change. The 2024 and 2025 changes landed in March, three months before a financial year ended. Court data by Indigenous status exists for only two years, so chart 4 is a before-and-after snapshot rather than a trend. Population rates use ABS projections after 2021, which will be revised after the 2026 Census.

Data and methods

All data is loaded directly from its publisher each time this report is built: Excel tables from the Crime Statistics Agency website, and CSV data from the ABS Data API. Processing was done in R (dplyr, tidyr) and all charts were made with plotly for R. Two independent data sources are used:

1. Crime Statistics Agency (CSA): bail data tables for Victoria Police, Corrections Victoria, the Magistrates’ Court and the Children’s Court, July 2016 to June 2026.
2. Australian Bureau of Statistics (ABS): quarterly estimated resident population by age and sex (ERP_Q); Aboriginal and Torres Strait Islander estimated resident population (ERP_ATSI); and Aboriginal and Torres Strait Islander population projections, Series B (ABORIGINAL_POP_PROJ).

Both sources publish these data under a Creative Commons Attribution 4.0 International licence, which permits reuse with attribution. The data are aggregate counts, and no individual can be identified.

Design notes

• Audience and approach. Written for general Victorian readers ahead of the state election. The narrative is author-driven, with reader-driven controls inside each chart (a hybrid approach; Segel & Heer, 2010).
• Encoding. Every key comparison uses position or length, the most accurate encodings (Cleveland & McGill, 1985). Colour carries a quantity only in the heatmap, where each cell also prints its value.
• Colour. Vermillion always means Aboriginal people and blue always means non-Indigenous people, a colour-blind-safe pair (Okabe & Ito, 2008). Grey and lighter tints de-emphasise context so the highlighted data stands out.
• Publication-ready. Every chart follows the data visualisation checklist (Evergreen & Emery, 2016): a headline that states the takeaway, horizontal text throughout, lines labelled directly and keys placed beside the data instead of separate legends, gridlines muted or removed where values are printed, and a source note under every chart.
• Integrity. Rates per 100,000 adults are used alongside counts, so population size doesn’t drive the comparison. Indexed lines replace a dual axis. Chart 5 lets readers change the baseline year to see how that choice affects the picture.
• Interactivity. Each control has one job: filter by sex (charts 1 and 4), change the encoding (rate ratio or rates, chart 2), reconfigure the view (number or share, chart 3; baseline year, chart 5), and show details on demand through hover (all charts; Yi et al., 2007).

Acknowledgements

I acknowledge the Traditional Owners of the lands on which this work was prepared, the Wurundjeri Woi-wurrung people of the Kulin Nation, and pay respect to Elders past and present.

GenAI acknowledgement. I used Claude Opus 5.5 (Anthropic) for the following limited purposes:

• suggesting refinements to my written work;
• reviewing spelling and grammar;
• suggesting references and other online sources relevant to the assignment, which I independently evaluated and verified; and
• providing help with technology relevant to the assignment, such as guidance on how to complete tasks in PowerPoint.

I reviewed all suggestions before deciding whether to use them. My research did not rely solely on GenAI, and I take responsibility for the final submission.