In 2016 the flat-screen television was the trophy of a Victorian burglary. Ten years later, almost nobody bothers. Televisions and video players recorded as stolen in Victoria fell 73%. Cameras fell 60%. Jewellery, once the 3rd most-stolen thing in the state, has slipped to 12th.
What replaced them is the strange part. The fastest-growing target in Victoria is not valuable at all. It is a strip of pressed aluminium worth about the price of a takeaway meal — and it points to a story not about greed, but about friction.
This report follows one question through ten years of Victoria Police property records: when something stops being worth stealing, what takes its place? The answer comes from 425,308 item-level theft records published by the Crime Statistics Agency, read against the prices Australian households actually pay.
Victoria Police record what was taken in every burglary and theft they attend. Group ten years of those records by item and a pattern appears that no single year would show: the mix changed completely.
Counts alone cannot tell this story honestly. The Crime Statistics Agency warns that fewer property-item reports are being submitted than a decade ago, which pushes every raw count downwards. So the measure used throughout this report is each item’s share of everything recorded stolen — a mix is unaffected by how many reports are filed.
Figure 1. What Victoria stopped stealing, and what replaced it
Each item's share of all property recorded stolen in Victoria, 2016-17 (hollow circle) to 2025-26 (filled circle). Jewellery lost 9.8 percentage points of the total; car parts and accessories gained 7.3 percentage points.
Try this: use the buttons to change where the item was taken from. Car parts barely move in the Residential view and jump in the Community view — the surge is happening on the street, not in the driveway. Hover any item for counts, shares and average recorded value.
Chart by Heet Chanchad. Source: Crime Statistics Agency Victoria (2026), Property items stolen, Table 01, year ending June 2026. Items keep the same order and the same x scale in all four views, so switching view never reorders the axis or quietly rescales it. Colour shows the direction of change within the view being displayed.
Jewellery is the clearest casualty. It was the 3rd most-stolen category in 2016-17 at 12.4% of everything taken; by 2025-26 it was 2.5%, a fall of 9.8 percentage points. Yet over the same decade the average recorded value of a stolen piece of jewellery rose 263%. Jewellery did not become less valuable. It became less available: residential burglary — the crime that reaches a jewellery box — fell 39% on the police count of offences.
Car parts and accessories went the other way, from 8th place to 2nd, and from 7.0% to 14.4% of everything stolen. Switch the buttons above and the reason shows itself: in community locations — streets, car parks, shopping centres — car-accessory items went from 9,787 in 2021-22 to 22,982 in 2025-26, while the same items taken from homes barely moved (6,163 to 8,227).
If theft were simply a search for value, the items rising up the list would be the expensive ones. They are not. Set each item’s frequency against what Victoria Police actually recorded it as being worth, and the pattern inverts.
Figure 2. The cheaper it is, the more of it gets stolen
Every item category, each year from 2016-17 to 2025-26. Horizontal position is the average value police recorded for one stolen item; vertical position is how much of all theft that item accounts for; bubble size is the number of items.
Try this: press Play the decade and watch two bubbles swap places. Jewellery drifts right and sinks; car parts climb while moving left, finishing at $117 — down from $161 in 2016-17. Drag the slider to any single year, or click a legend entry to isolate one group.
Chart by Heet Chanchad. Source: Crime Statistics Agency Victoria (2026), Property items stolen, Table 01. Bubble area is the number of items recorded stolen. Value is the amount recorded by police at the time of the report, not a market valuation. The horizontal axis is logarithmic because the categories span two orders of magnitude.
The most-stolen growth category in Victoria is also one of the cheapest things on the list. That only makes sense if the thief is not planning to sell it.
Three of the four fastest-rising categories are cheap and consumable: food (1.1% of all items stolen in 2016-17, now 3.0%), cigarettes and liquor (2.8% to 5.2%), and car parts. The police count of steal from a retail store offences — collected separately from property-item records — rose 144% between 2021-22 and 2025-26, which is the same shift seen from a different direction.
So what is the cheap thing being taken off cars in such volume? The Crime Statistics Agency publishes a table that answers it directly: offences in which a number plate was stolen.
Figure 3. Volume up, value down: the scissors at the heart of the story
Top: number-plate theft offences and car-accessory items recorded stolen in Victoria. Bottom: the average value police recorded for one stolen car accessory. Both panels cover the same reporting years.
Try this: hover anywhere to read both panels for the same year at once. Click a legend entry in the top panel to isolate one series.
Chart by Heet Chanchad. Source: Crime Statistics Agency Victoria (2026), Property items stolen, Tables 01 and 06. Table 06 counts offences in which a plate was stolen, not the number of plates. The two measures have different units, so they are drawn as two panels rather than on two y-axes, which can be scaled to make any two lines appear to move together.
Number-plate theft offences fell to 14,363 in 2021-22, then rose to 32,134 in 2025-26 — up 124% in four years, and roughly 88 a day. Set that against the 37,682 car-accessory items recorded in the same year and the category stops being a mystery: counting one offence as one item, plates account for around 85% of it, and for roughly 94% of everything it has gained since 2021-22. The car-accessory line in the panel above is, to a first approximation, a number-plate line.
A number plate has almost no resale value. Nobody fences a plate. What a plate buys is anonymity: a tank of petrol driven away from the pump, a toll gantry passed without a bill, a camera that photographs somebody else’s registration. Theft of this kind is not an end in itself. It is a means — a consumable input to the next offence. The police count of steal from a motor vehicle offences rose 66% over the same four years.
If plates are being taken to make a car untraceable, they should not be taken evenly across Victoria. They should cluster where there is enough traffic for an unfamiliar car to disappear into it. The map is blunt about where that is.
Figure 4. Plate theft stops at the edge of the city
Offences in which a number plate was stolen, per 10,000 residents, by Victorian local government area, 2025-26. A rate is used rather than a count so that a large rural shire and a small dense council can be compared fairly; on counts alone the populous councils would win by definition.
Try this: the three controls drive the map and the table together. Pick a police region, or drag a slider to keep only the areas above a given rate or growth, and the map fades everything that drops out while the table lists exactly what is left. Hover any area for its full numbers. The table is also the accessible route to the same data: it can be sorted, searched and read with a keyboard.
Chart by Heet Chanchad. Source: Crime Statistics Agency Victoria (2026), Property items stolen, Table 06, year ending June 2026; Australian Bureau of Statistics (2026), ERP by LGA (2025), for resident population at 30 June 2025; boundaries from ABS ASGS Edition 3 (2021). Base map © CARTO, data © OpenStreetMap contributors. All 79 Victorian local government areas recorded at least one such offence in 2025-26, so no area is missing from the map for want of data.
Every area at the dark end is in Melbourne, and they are not all the same kind of suburb. Maribyrnong has the highest rate in the state at 116.5 offences per 10,000 residents, with Greater Dandenong, Port Phillip, Brimbank close behind — inner-west, middle-south and inner-south, not one corridor. Casey records the most offences outright (2,754 in 2025-26) and has grown fastest of any substantial area, up 257% since 2021-22. Ten councils account for 49% of every plate-theft offence in Victoria.
The pale end is the whole of rural Victoria. Alpine recorded 4 such offences in 2025-26, a rate of 3.0 per 10,000 — about 39 times lower than Maribyrnong. That gap is not about rural policing. A stolen plate is worth having only where there is enough traffic to be anonymous in, and enough fuel pumps, toll roads and cameras to be anonymous from.
One explanation would tie all of this neatly together: thieves take whatever has become expensive. The ABS price indexes let that idea be tested directly, and it turns out to be half right — which is the more interesting answer.
Figure 5. What the country pays, and what Victoria loses
Six matched pairs. The solid blue line is what Australian households pay for that class of goods; the dashed orange line is how much of it is recorded stolen in Victoria. Both are indexed so that 2016-17 = 100, and all six panels share one scale. In the petrol panel the crime series is number-plate theft, not stolen fuel.
Try this: hover any point to read the index and year. Click a legend entry to hide one of the two lines and compare the six panels on the other.
Chart by Heet Chanchad. Source: Crime Statistics Agency Victoria (2026), Property items stolen, Tables 01 and 06; Australian Bureau of Statistics (2026), Consumer Price Index, Australia, original index numbers for the expenditure classes named in each panel. Quarterly prices are averaged into years ending 30 June to match CSA reporting. Co-movement is not causation, and the two panels that do not fit the pattern are shown alongside the four that do.
Four panels move together. Tobacco prices climbed and so did cigarette and liquor theft. Grocery prices climbed and food went from 1.1% of everything stolen to 3.0%, close to a tripling of its share. Petrol got dearer and plate theft followed it up. Electronics got cheaper almost every quarter for ten years, and theft of them collapsed.
Then there are tools. Hardware prices rose like everything else, yet tool theft fell by about half. And jewellery, from Chapter 1, became far more valuable while being stolen far less. Price alone does not explain either.
What the exceptions share is difficulty. Tools are serial-numbered and sold into a trade that knows its suppliers. Jewellery sits inside houses, and breaking into houses has become rarer and riskier.
Put the pattern and the exceptions together and a single rule fits all six panels: theft migrates towards whatever is easiest to convert. Not the most valuable thing — the thing with the least friction between taking it and using it. A television still records about $1,106 when one is taken, but it is bulky, serial-numbered, and the second-hand market for it has all but gone — so it stays on the wall. A gold ring is worth more than ever and sits behind a locked door. A supermarket steak, a packet of cigarettes and a number plate on a street in Casey are worth almost nothing each, and all three turn into value the moment they are picked up.
Victoria’s recorded property crime is routinely read as a thermometer of disorder. Read item by item, it looks much more like an economic indicator. Three things follow.
A falling count is not all good news. Fewer property items are recorded stolen than ten years ago, and the categories that emptied out are the resale ones — electronics, cameras, jewellery — which stopped being worth the trouble. The categories tied to the weekly shop went the other way, and on the police count of offences, rather than of items, theft from a retail store rose 144% in four years.
The cheapest offences carry the highest leverage. 32,134 plate-theft offences a year generate almost no recorded property loss, and the Crime Statistics Agency’s own notes caution that property records are incomplete — so a measure built on dollars lost would miss this almost entirely. A measure built on what enables the next offence would not.
Prevention has to follow the friction. Anti-theft number plate screws cost a few dollars. On this evidence they stand in front of an offence that has more than doubled in four years and that feeds the next one. Raising the friction on the cheapest thing on the list is not a small intervention; on this reading it is the whole point.
What a “property item” is. The Crime Statistics Agency counts one item per property type recorded in an offence, not the quantity: 50 litres of stolen petrol and one stolen jerry can both count as one petrol item. Figures here are therefore counts of records, never volumes or weights.
Why shares, not counts. The agency states that the number of property-item reports submitted for theft and burglary offences has fallen over time, and that some offences carry no property type at all. That depresses every raw count. The mix between items is far more robust to it, so share of all items is the measure used for every composition claim. Where an absolute count is quoted, it comes from the recorded-offence series, which is collected independently — this is why plate theft, retail theft, steal-from-a-motor-vehicle and residential burglary are all cited from offence counts rather than item records.
“Other / unclassified” is large. It is 18% of items in 2025-26 and is shown rather than hidden, because leaving it out would inflate every other share.
Recorded value is not market value. The dollar figures are what police recorded at the time of the report. They are useful for comparing items against each other and over time; they are not valuations.
Plate offences are offences. Table 06 counts offences in which at least one plate was stolen, not plates. The comparison in Chapter 3 counts one offence as one item, which is an approximation and is described as such.
Years and geography. All CSA years end 30 June, so 2025-26 means July 2025 to June 2026. Rates use ABS estimated resident population at 30 June 2025, the latest available, against 2025-26 offences. Boundaries are ABS ASGS Edition 3 (2021), which still names Merri-bek “Moreland”; the join is reconciled in code and asserted to leave no area unmatched.
Prices. CPI series are original (not seasonally adjusted) index numbers for Australia, averaged from quarters into years ending 30 June. National rather than Melbourne figures are used because the current CPI series publishes expenditure-class detail at the national level only, and the superseded series that carries Melbourne detail ends a year before the crime data — so using it would have compared two different decades. Prices for these goods are set in national markets, which makes the substitution a minor one, but it is a substitution and the axis labels say so.
Correlation. Chapter 5 shows co-movement between prices and theft. It does not establish that one causes the other. The two categories that do not fit the pattern are shown in the same figure as the four that do, rather than left out.
Colour is assigned by the job it does. Direction of change uses a diverging pair (two poles that read as opposites, with a neutral grey midpoint for items that barely moved); the map uses a single blue hue running light to dark for magnitude; multi-series charts draw from a fixed categorical order that is never cycled. Each set was checked for colour-vision-deficiency separation, a normal-vision distinctness floor, and contrast against the chart surface.
Colour never carries meaning alone. Figure 1 names every item on the axis and labels both end-points in the legend; Figure 3 prints its end values; Figure 5 pairs colour with line style; Figure 4 is accompanied by a sortable, searchable table of the same numbers, which is also the accessible route to the map for screen-reader and keyboard users.
No figure uses two y-axes. Where two measures have different units (Figure 3) they are drawn as two stacked panels, and where many series share a unit (Figure 5) they are indexed to a common base — both in preference to a second axis, which can be scaled to manufacture any correlation the author wants.
Interactive charts are an addition to the argument, never the whole of it. Every finding is stated in the running text, so the report can be read without touching a control. Interaction is reserved for the three places where it carries information a static figure cannot: the location switch in Figure 1, the decade animation in Figure 2, and the linked map, filters and table in Figure 4.
Data are published by the Crime Statistics Agency Victoria under a Creative Commons Attribution 4.0 International licence, and by the Australian Bureau of Statistics under Creative Commons Attribution 4.0 International. Both agencies are acknowledged as the source of all figures in this report. Neither endorses this analysis.
Generative AI. Anthropic’s Claude (Opus 5) was used as
a coding and editing assistant during this assignment. It was used to
help structure the R Markdown document, to debug plotly,
crosstalk and leaflet code, to suggest wording
for explanatory passages, and to check the colour palettes against
colour-vision-deficiency and contrast thresholds. The story, the choice
of datasets, the analytical decisions (including the decision to analyse
shares rather than counts), the interpretation and the conclusions are
the author’s own. All outputs generated with assistance were verified
against the source data, and every figure quoted in the text is computed
in the document from the loaded data rather than typed in by hand. This
acknowledgement follows the RMIT Library guidance on acknowledging and
referencing artificial intelligence.
Tools. Built in R with rmarkdown,
dplyr, tidyr, ggplot2,
plotly, leaflet, sf,
rmapshaper, crosstalk and DT.
Base map tiles are CARTO Positron, with data from OpenStreetMap
contributors.
Australian Bureau of Statistics. (2021). Australian Statistical Geography Standard (ASGS) Edition 3: Local government areas [Data set]. ABS geospatial services. https://geo.abs.gov.au/arcgis/rest/services/ASGS2021/LGA/MapServer
Australian Bureau of Statistics. (2026). Consumer Price Index (CPI) [Data set]. ABS Data API. https://data.api.abs.gov.au/rest/data/ABS,CPI,2.0.0/
Australian Bureau of Statistics. (2026). ERP by LGA (2025), age and sex, 2001 to 2025 [Data set]. ABS Data API. https://data.api.abs.gov.au/rest/data/ABS,ABS_ANNUAL_ERP_LGA2025,1.0.0/
Anthropic. (2026). Claude (Opus 5) [Large language model]. https://claude.ai
Cairo, A. (2016). The truthful art: Data, charts, and maps for communication. New Riders.
Cheng, J., Schloerke, B., Karambelkar, B., & Xie, Y. (2026). leaflet: Create interactive web maps with the JavaScript “Leaflet” library [R package]. https://rstudio.github.io/leaflet/
Crime Statistics Agency. (2026). Data tables: Property items stolen, Victoria, year ending June 2026 [Data set]. Victorian Government. https://www.crimestatistics.vic.gov.au/crime-statistics/latest-victorian-crime-data/download-data
Crime Statistics Agency. (2026). Data tables: Recorded offences, Victoria, year ending June 2026 [Data set]. Victorian Government. https://www.crimestatistics.vic.gov.au/crime-statistics/latest-victorian-crime-data/download-data
Crime Statistics Agency. (2026). Explanatory notes. Victorian Government. https://www.crimestatistics.vic.gov.au/about-the-data/explanatory-notes
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. (2026). R: A language and environment for statistical computing (Version 4.6.1) [Computer software]. R Foundation for Statistical Computing. https://www.R-project.org/
Segel, E., & Heer, J. (2010). Narrative visualization: Telling stories with data. IEEE Transactions on Visualization and Computer Graphics, 16(6), 1139–1148. https://doi.org/10.1109/TVCG.2010.179
Sievert, C. (2020). Interactive web-based data visualization with R, plotly, and shiny. Chapman & Hall/CRC. https://plotly-r.com
Wickham, H. (2016). ggplot2: Elegant graphics for data analysis (2nd ed.). Springer. https://doi.org/10.1007/978-3-319-24277-4
World Wide Web Consortium. (2023). Web content accessibility guidelines (WCAG) 2.2. https://www.w3.org/TR/WCAG22/
Xie, Y., Allaire, J. J., & Grolemund, G. (2018). R Markdown: The definitive guide. Chapman & Hall/CRC. https://bookdown.org/yihui/rmarkdown/
Five interactive visualisations, two independent data sources, all figures built in R and published to a single URL. Data read directly from the Crime Statistics Agency and the ABS at build time; 27 October 2026.