Living around Prahran made me interested in how people talk about crime in the area. Prahran and South Yarra are often discussed together, but I wondered whether the recorded offences in these places actually followed the same patterns. Instead of trying to label a suburb as safe or unsafe, I wanted to look at something more specific: which offences are recorded most often, how have they changed, and what differences appear between neighbouring suburbs?
I used ten years of recorded-offence data for Stonnington, covering the annual periods ending June 2017 to June 2026. The story starts with the council-wide picture, moves into six suburbs, and finishes by looking more closely at Prahran.
How to explore the charts: Hover over points or sections for details, click legend items to focus on a category, and use the chart toolbar to zoom or reset the view.
The first thing I noticed was how far property and deception offences sit above the other categories. There were 7,606 recorded offences in this division in 2017, compared with 9,480 in 2026. That is an increase of approximately 24.6% over the period. The count was even higher in 2025, at 10,078, before falling slightly in the latest year.
This doesn’t mean that every type of crime increased in the same way. The lines for other divisions are much lower and follow different patterns. It made me want to look beyond Stonnington’s total and see whether nearby suburbs have similar offence profiles.
Property and deception offences are the largest category in every suburb I compared, but they do not make up exactly the same proportion everywhere. In 2026, this category accounted for approximately 67.6% of Prahran’s recorded offences and 70.4% of South Yarra’s. Its share was higher in Toorak (77.4%) and Armadale (78.1%).
This is why I chose to use percentage bars rather than only raw totals. South Yarra recorded more offences overall than some of the other suburbs, but a larger total does not automatically mean the mix of offence types is the same. These comparisons describe the composition of recorded offences, not which suburb has the greatest personal safety risk.
After seeing how much property crime contributed to the totals, I wanted to know what was inside that category. Theft stands out across all six suburbs. Burglary or break and enter, property damage, deception and arson account for smaller shares, and their proportions are not identical from suburb to suburb.
I found this more useful than treating every property offence as one thing. Even when the overall category looks similar, the activities recorded within it can differ. The colours show the share within each suburb’s property offences; they do not show the probability of experiencing theft or burglary.
One particular type of theft caught my attention: items stolen from motor vehicles, rather than the vehicles themselves. The lines show relative change, so every suburb starts at 100 in 2017. By 2026, South Yarra had gone from 297 to 678 recorded offences, an increase of approximately 128.3%. Prahran increased from 224 to 480 (114.3%), while Windsor rose from 107 to 225 (110.3%).
The contrast with Toorak, where the count was 233 in 2017 and 236 in 2026, is striking. Malvern actually fell slightly, from 166 to 157. These differences were less obvious when I was looking only at the broad property-offence category. Because this graph uses an index, it is important to check the actual counts as well: a large percentage increase can start from a relatively small number.
I wanted to finish by returning to Prahran. The clearest change is theft from motor vehicles. It made up 8.11% of Prahran’s recorded offences in 2017, but 16.9% in 2026. The number of recorded offences in that subgroup rose from 224 to 480.
Other categories moved in the opposite direction. Obtaining a benefit by deception fell from 13.0% to 4.82% of the suburb’s total, with recorded counts declining from 358 to 137. Drug possession also fell, from 9.67% to 5.45%, or 267 to 155 offences. These are changes in both recorded counts and their share of the total; the graph itself focuses on the shares.
For me, this was the most important part of the story. Prahran did not simply experience one uniform change in “crime.” Different offence types moved in different directions, and the overall total hides those details.
Before exploring the dataset, it would have been easy to focus on which suburb recorded the most offences. The five charts gave me a more useful way to understand the numbers: first by looking at the long-term trend, then comparing the offence mix, and finally investigating a specific type of theft and the changes within Prahran.
The strongest finding is not that one suburb is safer or more dangerous than another. It is that the kind of offence, the location and the period being compared all matter. Recorded-offence figures can start a more informed conversation about an area, but they cannot tell us everything about the experiences of people living or spending time there.
I used publicly available Victorian recorded-offence data from the Crime Statistics Agency (CSA), accessed through a versioned Stonnington JSON file republished by National Digital / publicdata.au. The republication reformats the original government dataset and is not an independent crime survey.
I imported the data into R, checked the year and suburb fields, and
used dplyr to calculate yearly totals, offence-category
totals, subgroup counts and percentage shares. I created the
visualisations with ggplot2 and made them interactive with
ggplotly() from the plotly package. The R
analysis file contains the data-processing and chart code.
There are important limits to this analysis. Recorded offences are not the same as unique incidents, victims or offenders, and not every offence is reported to police. Raw suburb totals can be affected by the number of residents, visitors, businesses, reporting patterns and police activity. The dataset does not supply a population or visitor denominator for these charts. The data therefore cannot be used to calculate personal crime risk, identify causes, or claim that nightlife led to any of the patterns shown. Each year refers to a 12-month period ending in June, not a calendar year.
Crime Statistics Agency. (2026). Data tables – recorded offences [Data set]. Victorian Government. https://discover.data.vic.gov.au/dataset/data-tables-recorded-offences
National Digital. (2026). Recorded offences by suburb and offence, Victoria (version 2026-09-24) [Reformatted data set]. publicdata.au. https://publicdata.au/d/vic-recorded-offences-by-suburb/v/2026-09-24/
I used ChatGPT (OpenAI) to explain some R code,and improve the structure and wording of this report. I ran the charts in Posit Cloud and checked the reported findings against the data outputs. I am responsible for reviewing the final report and acknowledging the assistance in my Cadmus submission.