Why St Kilda?

St Kilda has a reputation, and most people who have spent time in Melbourne have heard it. When I started looking at the Victorian crime statistics I assumed the numbers would back that reputation up, mostly through violent crime. The data did not really say that.

What it showed instead was a rise in recorded theft, and most of that rise is shop theft. This report follows that through five charts. The first two place St Kilda’s council area among the rest of Melbourne. The third and fourth look at what has changed. The fifth checks whether St Kilda is doing something different from the rest of Victoria.

Using the charts. You can hover over anything for exact numbers, drag to zoom, and double-click to reset. Chart 1 has a play button and two zoom buttons. Charts 2 and 3 have buttons above the plot. Chart 5 has a search box: type a council’s name, or click a bubble, and it lights up in all four panels.

Crime figures are published by council area, not by suburb. St Kilda sits inside the City of Port Phillip, so Charts 1, 2, 3 and 5 are about Port Phillip as a whole, and Chart 4 uses the suburb-level tables. All the crime data come from the Crime Statistics Agency (CSA, 2026).

Chart 1: Where Port Phillip sits

Source: Crime Statistics Agency (2026), Table 01; council boundaries from the Australian Bureau of Statistics (ABS, 2025). Darker shading means more recorded offences per 100,000 residents. Press Play or drag the slider to move through the years, and use the buttons above the map to zoom.

Port Phillip is the outlined council on the bay. Since 2017 it has had the 3rd-highest rate of the 31 councils in Greater Melbourne in every single year. The City of Melbourne is the darkest by a clear margin, which makes sense for an area with the CBD and a huge daytime population. Port Phillip sits in the next group down and stays there as the years change.

So the reputation is not new. St Kilda’s area has ranked near the top of Melbourne for as long as this data goes back, which means the interesting question is what has changed within that ranking.

Text description of Chart 1 (accessibility) A map of the 31 councils of Greater Melbourne shaded by offences per 100,000 residents, animated from 2017 to 2026. The City of Melbourne is always the darkest. Port Phillip, outlined in black, is in the next-darkest group and has the 3rd-highest rate of the 31 councils in Greater Melbourne in every single year. In the year ending June 2026 its rate was 14,100 offences per 100,000 residents.

Chart 2: The rise came in the last two years

Source: Crime Statistics Agency (2026), Table 01. The coloured lines are neighbouring inner-city and bayside councils, and the dashed black line is the median of all 31 Greater Melbourne councils. The buttons switch between the index view and the raw rate. The index view does not start at zero because it shows change over time, not size.

In the index view every council’s 2017 rate is set to 100, so a line at 110 means a rate ten per cent above 2017. That makes it easier to compare change when the councils start from very different levels. The second button shows the raw rates instead.

Port Phillip dropped to 94 in 2024 and then climbed to 113 by 2026, a rise of about 20 per cent in two years. The median council in Greater Melbourne finished 2026 at 98, so it is close to where it started. Stonnington (118) went up even more than Port Phillip, while Glen Eira and Bayside are below 100. The increase is in the inner city and not across Melbourne generally.

That made me want to know which offences are behind it.

Text description of Chart 2 (accessibility) A line chart of recorded offence rates from 2017 to 2026 for Port Phillip, five comparison councils and the Greater Melbourne median. Port Phillip’s index falls to 94 in 2024 and rises to 113 in 2026. The Greater Melbourne median ends at 98.

Chart 3: What changed

Source: Crime Statistics Agency (2026), Table 02 (Port Phillip only). The bars show the ten offence types that changed the most between 2022 and 2026. Theft is red and everything else is grey. The buttons change the starting year.

Between 2022 and 2026 recorded offences in Port Phillip went from 12,150 to 16,196, an increase of 4,046. Theft accounts for 3,133 of that, about 77 per cent, rising from 4,718 to 7,851.

Violence does not explain the change. Assault and related offences were 1,096 in 2022 and 1,138 in 2026, and because the population grew over those years the assault rate actually went down a little. Starting from 2017 or 2024 instead gives the same answer, with theft at the top.

Theft is a broad category, so the next chart splits it up.

Text description of Chart 3 (accessibility) A horizontal bar chart of the change in recorded offences in Port Phillip for the ten offence types with the largest change. Theft has by far the longest bar, increasing by 3,133 offences between 2022 and 2026. The other bars, in grey, are much shorter.

Chart 4: Inside theft, shop theft stands out

Source: Crime Statistics Agency (2026), Table 03 (suburb level, Port Phillip only). Every panel uses the same vertical scale so they can be compared directly. These are counts and not rates, because the data has no suburb populations.

Each panel is one type of theft. The coloured lines are the three suburbs with the most shop theft in 2026, and the grey line is everywhere else in Port Phillip.

The top-left panel is the important one. Shop theft in Port Phillip was 224 in 2022 and 1,579 in 2026, and most of the growth came late: 454 in 2024, 929 in 2025 and then 1,579. The suburbs do not move together. South Melbourne more than doubled in 2025 (181 to 425), while Balaclava (94 to 432) and St Kilda (125 to 296) jumped in 2026. I would not call this one suburb having a bad year. Stealing from motor vehicles, the next biggest type, has grown as well but far more slowly.

I wanted to know whether this was particular to Port Phillip or part of a wider trend, which is what the last chart looks at.

Text description of Chart 4 (accessibility) Four small line charts of theft offences in Port Phillip from 2017 to 2026, one for each type of theft, with separate lines for the three suburbs with the most shop theft and for all other suburbs. Shop theft rose from 224 offences in 2022 to 1,579 in 2026. All panels use the same vertical scale.

Chart 5: Is St Kilda different from the rest of Victoria?

Source: Crime Statistics Agency (2026), Table 02 and Table 01 (population worked out from the rates). Each bubble is a Victorian council and bubble size shows population. Anything above the dashed line has a higher rate in 2026 than in 2022. Councils with no recorded figure in either year are left out. Port Phillip is always shown as the red diamond. Click a bubble or use the search box to highlight a council in all four panels, and double-click to clear.

Shop theft is the first panel. Almost every bubble sits above the dashed line: of the 75 councils with data, 93 per cent had a higher shop-theft rate in 2026, and the typical council’s rate multiplied by about 2.3. So St Kilda did not start this. It is part of a Victoria-wide rise.

Port Phillip stands out from the pack, though. Its rate multiplied by 6.4, from 216 to 1,375 per 100,000, which moved it from 33rd to 3rd of the 75 councils. That multiple is roughly 2.8 times the typical one. Now compare the assault panel: there the diamond sits close to the line, with a factor of 0.94 against 1.05 for the typical council. Violence is not where the movement is.

Text description of Chart 5 (accessibility) Four scatter plots, one each for shop theft, theft from a motor vehicle, motor vehicle theft and assault, comparing every Victorian council’s 2022 and 2026 rate with a dashed no-change line. In the shop theft panel nearly all councils are above the line and Port Phillip, a red diamond, is far above it. In the assault panel Port Phillip is close to the line.

What I take from this

I went in expecting violence to be the story, and it is not. Port Phillip has ranked high for years, but what has actually changed in the last two years is theft, and shop theft in particular. About 77 per cent of the extra offences since 2022 are theft. Shop theft has risen across Victoria, but in Port Phillip it has risen much faster than in the typical council, while assault has stayed level.

What the data cannot tell me is why. More theft is one explanation. Another is that more of it is being reported or recorded, for example if retailers or police now handle shop theft differently. Separating the two would need other information, such as retailer reports or police practice, and that is where I would look next.

Limitations

  • Recorded offences are not all crime. The CSA counts offences that were reported to or detected by Victoria Police and then recorded. A change in reporting looks the same as a change in behaviour.
  • Rates use resident population. Visitors and workers are not in the denominator, so places with many of them, like the City of Melbourne and parts of Port Phillip, can look worse per resident than they are for people who live there.
  • Suburb figures are counts. Chart 4 cannot be compared by population size because the data has no suburb populations.
  • Missing values. Some councils have no figure for an offence type in one of the two years. They are left out of Chart 5 rather than guessed.
  • Boundaries are approximate. The ABS builds council boundaries from mesh blocks for statistical use, so the shapes are a close approximation of the official ones.

Data and method notes

  • Source 1: Crime Statistics Agency (CSA). Data tables: Recorded offences by LGA, year ending June 2026 (Excel workbook; Tables 01 to 03 used). It reports the years ending June 2017 to June 2026.
  • Source 2: Australian Bureau of Statistics (ABS). Local Government Areas 2025 digital boundary file (ASGS Edition 3), used only to draw the map in Chart 1.
  • Cleaning. Rows that are not councils (“Total”, “Unincorporated Vic”, “Justice Institutions and Immigration Facilities”) were removed. Four council names that differ between the two sources (Latrobe, Bayside, Kingston and Colac Otway) were matched by hand, and extra spaces were trimmed.
  • Population. The file gives rates but not populations, so population was calculated as offence count divided by rate per 100,000, multiplied by 100,000.
  • Greater Melbourne is the list of 31 councils in the code, including Mornington Peninsula and Yarra Ranges.
  • Map. Boundaries were simplified to a 250 m tolerance so the page loads quickly.
  • Design choices. The palette is colour-blind safe (Okabe & Ito, 2008). Vermillion is kept for Port Phillip and theft only, following the idea of using colour with a purpose (Few, 2008). Position and length are used for comparisons wherever possible because people read them more accurately than area or colour (Cleveland & McGill, 1985). The charts were built in R with plotly (Sievert, 2020), ggplot2 (Wickham, 2016) and sf (Pebesma, 2018).

Acknowledgements

I developed the data analysis, R code, visualisations, interpretation and report content myself. I used Claude (Anthropic) only for minor editorial touch-ups, such as checking wording, grammar and clarity. Claude was not used to create the analysis, write the R code, produce the visualisations or determine the findings. I reviewed the final work and take responsibility for the analysis, interpretation, references and submitted report.

References

Anthropic. (2026). Claude [Generative AI tool]. https://claude.ai

Australian Bureau of Statistics. (2025). Local Government Areas: Australian Statistical Geography Standard (ASGS) Edition 3 [Digital boundary file]. https://www.abs.gov.au/statistics/standards/australian-statistical-geography-standard-asgs-edition-3/jul2021-jun2026/access-and-downloads

Cleveland, W. S., & McGill, R. (1985). Graphical perception and graphical methods for analyzing scientific data. Science, 229(4716), 828-833.

Crime Statistics Agency. (2026). Data tables: Recorded offences by LGA, year ending June 2026 [Data set]. State of Victoria. https://discover.data.vic.gov.au/dataset/data-tables-recorded-offences

Few, S. (2008). Practical rules for using color in charts. Perceptual Edge. http://www.perceptualedge.com/articles/visual_business_intelligence/rules_for_using_color.pdf

Okabe, M., & Ito, K. (2008). Color universal design (CUD): How to make figures and presentations that are friendly to colorblind people. https://jfly.uni-koeln.de/color/

Pebesma, E. (2018). Simple features for R: Standardized support for spatial vector data. The R Journal, 10(1), 439-446. https://doi.org/10.32614/RJ-2018-009

R Core Team. (2025). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.R-project.org/

Sievert, C. (2020). Interactive web-based data visualization with R, plotly, and shiny. Chapman and Hall/CRC. https://plotly-r.com

Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. Springer-Verlag.