How we tested what drives credit and house prices

Every figure here is computed from official, openly licensed data by code first tested on invented data whose answers are known. Each model was fitted on one period, or one set of councils, and scored on another it had not seen.

Part A. Capacity to repay and credit

How the chain works

Inflation above the RBA’s target leads it to raise the cash rate. Lenders pass that to variable mortgage rates, and repayments rise. Inflation also raises living costs, so less of the same income is left for the loan. Unemployment takes the income away altogether. Whether a missed payment becomes a loss depends on equity: a borrower with equity can sell, and one in negative equity cannot. Economists call needing both an income shock and negative equity the double trigger.

What the RBA’s research found

• Bergmann (2020), “The Determinants of Mortgage Defaults in Australia – Evidence for the Double-trigger Hypothesis”, RDP 2020-03. A 4-point rise in a region’s unemployment rate roughly doubles a loan’s risk of entering arrears; loans deep in negative equity are about six times as likely to go from arrears to foreclosure.
• Read, Stewart and La Cava (2014), “Mortgage-related Financial Difficulties: Evidence from Australian Micro-level Data”, RDP 2014-13. Arrears rise with the loan-to-valuation ratio at origination, most above 90%; households paying more than half their income in debt service miss payments more often.
• Kearns, Major and Norman (2020), “How Risky is Australian Household Debt?”, RDP 2020-05. In a severe downturn, household debt at risk rises from 0.8% to 2.1% of housing debt, counting a default only where a household runs short and owes more than 90% of its home’s value.

What the public data shows

Rates pass through. From March 1997 to September 2026, each point on the cash rate moved the standard variable rate 0.98 points in the same quarter (R² 0.96, 119 quarters). The rate existing borrowers pay on average, published since 2019, moved 0.87 points.

The buffer was used up. In December 2021 a new variable loan cost 2.6% and was tested at 5.6%, the rate plus APRA’s 3-point buffer. By July 2026 the average borrower paid 6.2%, past that test (RBA table F6), while consumer prices rose 21.5% from December 2021 to June 2026 (ABS via RBA table G1).

Arrears follow unemployment now and rates a year later. The lags were chosen on March 2019 to December 2023 only, from unemployment 0–4 quarters back and the rate 0–8. The model put 0.10 points on the non-performing share for each point of unemployment (t 13.2) and 0.024 for each point on the rate 4 quarters earlier (t 2.1). On the ten quarters since, its average miss was 0.13 points, against 0.17 for assuming the last value stays. It saw the rise coming but not its size: from December 2023 to March 2025 the share rose 0.22 points, and the model 0.08. Without the deferral quarters of 2020–21 the estimates barely move.

The figures

Housing loans non-performing, actual and model

Quarter	Actual	Model
March 2019	0.87%	0.91%
June 2019	0.93%	0.93%
September 2019	0.92%	0.93%
December 2019	0.90%	0.92%
March 2020	0.94%	0.93%
June 2020	1.11%	1.09%
September 2020	1.06%	1.10%
December 2020	1.01%	1.07%
March 2021	1.04%	0.99%
June 2021	1.01%	0.91%
September 2021	0.91%	0.85%
December 2021	0.82%	0.85%
March 2022	0.82%	0.79%
June 2022	0.78%	0.77%
September 2022	0.71%	0.74%
December 2022	0.68%	0.73%
March 2023	0.72%	0.75%
June 2023	0.76%	0.75%
September 2023	0.81%	0.80%
December 2023	0.86%	0.84%
March 2024	0.96%	0.86%
June 2024	1.04%	0.88%
September 2024	1.06%	0.89%
December 2024	1.05%	0.89%
March 2025	1.08%	0.90%
June 2025	1.07%	0.91%
September 2025	1.04%	0.92%
December 2025	0.99%	0.91%
March 2026	0.99%	0.91%
June 2026	1.01%	0.92%

Equity decides the loss. The non-performing share climbs with the loan-to-valuation ratio: 0.58% under 60%, 6.70% at 95% or more (APRA).

The figures

Non-performing loans by loan-to-valuation band

Loan-to-valuation ratio	Non-performing, average	Latest quarter
Under 60%	0.58%	0.62%
60% to 80%	0.73%	0.79%
80% to 90%	1.54%	2.00%
90% to 95%	1.93%	Not published
95% or more	6.70%	5.88%

Inflation works through rates. Added to the arrears model, inflation a year earlier is not distinguishable from nothing (t −1.5). And while the RBA did respond to inflation from 2015 (t 4.6), a rule fitted to 1993–2014 forecast next year’s cash rate worse than assuming no change: an average miss of 1.09 points against 0.66, with the direction right in 5 of 29 moves.

Where the data is too thin. Before 2019 the only public series is the impaired and past-due share of all loans by banks, credit unions and building societies, business and household together (2004–2021). It does not track unemployment (correlation 0.15), and a model fitted to 2013 missed by 0.51 points afterwards against 0.38 for no change. Housing-only data covers one cycle, 30 quarters.

Ahead. On the RBA’s August 2026 forecast of unemployment (4.7% by December 2027) and the rate where it stands, the model fitted to every quarter puts the non-performing share near 1.09% by then, give or take 0.13 points.

Part B. House prices

What the RBA’s housing model says

• Saunders and Tulip (2019), “A Model of the Australian Housing Market”, RDP 2019-01. A sustained one-point fall in interest rates raises prices by 17% in the long run at a 6% user cost, and 28% at 3½%; a cut expected to last three years raises them 8% after two years. Each 1% added to the dwelling stock by supply lowers the cost of housing by 2½%, and immigration raised rents in the late 2000s.

Melbourne’s and country Victoria’s own history

Real growth in Melbourne’s yearly median house price, 1989–2025, against the change in the mortgage rate that year and the year before, and in unemployment. The signs are the RBA’s: −3.0% for each point of unemployment (t −2.0), −1.1% and −1.4% for each point on the rate this year and last. The effects are smaller and less certain than the RBA’s. Fitted to 1989–2010 and given the actual rates and unemployment of 2011–2025, it missed by 4.8% a year against 5.2% for average growth (country: 3.4% against 3.8%).

Forecasting them first is the hard part. From each September since 2005, with the rate held and the RBA’s forecasts where its archive has them, the model missed Melbourne’s two-year change by 7.4% on average, against 6.8% for average growth: no better. Income growth shows up over the full 37 years (t 4.3), but with income and population added, the model fitted to 2010 missed by 9.4% a year afterwards, so the forecasts leave them out.

What sets a suburb’s price

746 Victorian suburbs’ 2025 median house prices, on what the Census, SEIFA, the ABS’s boundaries and the timetable say about each. Tested ten times, each time on councils the model had not seen: it explains 80% of the differences, and misses a typical suburb by 13%.

Effect: the change in a suburb’s median with the others held equal. Lost if left out: the fall in the share explained, on councils held out. Alone: the share one feature explains by itself. Socio-economic advantage and income move together, so whichever is in the model carries the other’s weight.

Proxies for what has no open measure. Each added to the model on the suburbs where it exists, from the Valuer-General’s council figures and the Census:

• Council median house block size (Valuer-General 2025 (council level)): +19.1% when it doubles, t 7.3, adds 0.8 points to the share explained.
• Council median shop price per m² (proxy) (Valuer-General 2025 (council level)): +0.1% when it doubles, t 0.1, adds nothing to the share explained.
• Share of residents working in mining (proxy) (Census 2021): +2.4% per point, t 1.8, adds nothing to the share explained.

No open, suburb-level measure of commercial valuation or natural resources exists. These proxies are the closest the open data allows, and neither adds anything. Land does: at council level, bigger house blocks go with dearer houses.

What sets a suburb’s growth

The same features, measured at the start, fitted to growth from 2016 to 2021 and tested on 2021 to 2025 (Valuer-General Victoria). In sample they explain 47%. On the later years they beat a guess of average growth for every suburb by 4%, and four of them (median household income, share of homes that are separate houses, on the coast or the bay, outside Greater Melbourne) changed sign. Most of the change was the market’s: average growth fell from 8.3% a year to 2.2%. Growth in the first window says nothing about the second (correlation −0.04).

One feature did better: cheaper suburbs outgrew dearer ones in 2021–2025, and the starting price alone beat average growth by 25% (Valuer-General Victoria). That was found after seeing the test, so it was tried again on every pair of back-to-back three-year windows: it helped in three of six. Fitted to 2014–2017, when dearer suburbs led, it made 2017–2020 forecasts 39% worse than average growth. Which kind of suburb leads changes with the cycle.

Part C. The example forecasts

Illustrative only. Not a valuation, not a prediction to rely on, and not advice. The ranges themselves are in the piece.

How the ranges are made

1. The market. Melbourne’s or country Victoria’s model above, from September 2026: the standard variable rate held at 8.77%, and unemployment and inflation on the RBA’s August 2026 forecasts (4.5% and 3.6% for December 2026). The RBA’s own rate path comes from market data it licenses, so none is assumed.
2. The suburb. Its 2025 median moves with its market. Nothing tested above tells two suburbs of the same kind apart over two years, so the method does not try.
3. The range. The 5th to 95th percentile of the method’s past misses: the market’s, from every September since 2005, plus how far suburbs of the same kind strayed from their market since 2014. Tested on years left out, these ranges held 79% of two-year outcomes and 74% of three-year ones.
4. The suburbs. Picked by rule, not result: in each role, the most populous suburb with an unflagged median every year since 2014 (how each figure was made).

How wrong it has been. From each September since 2005, the method missed Melbourne’s market by 4.7% in the year under way, 7.4% two years out and 8.1% three years out, no better than assuming average growth (6.8% two years out). The example suburbs’ own two-year misses since 2016 averaged 8% to 17%.

Early 2026 sales. Melbourne’s preliminary median for early 2026, from 8,332 sales, is −6.1% on 2025, below the bottom of the method’s range for the year (−3.4% to +13.4%). Country Victoria’s, +3.8%, is inside it. The RBA wrote in August that “housing prices have declined noticeably”. The model knows the mortgage rate is 0.48 points above its 2025 average; it does not know how buyers took the rises. With the RBA model’s rate effect in place of this one, that would take about 2.7% more off Melbourne’s prices over two years. And the last time rates rose fastest, in 2022, the method’s midpoint from September 2022 for Melbourne’s 2023 median was +23% on 2021; it came in at −1%.

So the midpoints are not useful forecasts; the ranges are the result. Two years out, a Melbourne suburb’s range runs about 53 points from end to end, and the method has been no better than average growth.

Part D. New stations and new centres

The founder named two assumptions on 4 October 2026, and this part tests them as they were meant: that a suburb gains when it gets a new station — when the line is planned, the route confirmed, the station built or opened — and that it gains when a shopping centre, a business hub, a precinct or a new city centre is announced nearby. Distance to a station that already exists, which Part B measured, is not the same question.

How it was tested

1. The projects. seven rail projects and seven commercial ones in Melbourne, each with up to four milestones dated from official Victorian Government pages: announced, the route named or the plan approved, construction begun, opened. A milestone counts from the year it falls in if by 30 June, otherwise the next.
2. Near. A suburb is near a project if the station or site stands inside it (the ABS’s boundaries), or its centre is within 1.5 km — most of a typical suburb’s homes within a walk or a short ride. For SRL East, whose stations have no published point yet, a suburb is near if its centre lies in a declared planning area.
3. Far. Each near suburb is compared with the five suburbs most like it — distance to the CBD, the price the year before the announcement, and growth over up to three years before that — among the 191 suburbs at least 5 km from every project here and from the 14 other stations opened in Greater Melbourne since 2006 or planned that we dated (Department of Transport and Planning).
4. The gap. The near suburb’s change in its median house price since the year before the announcement, less the average change of its matches. Prices from 2006 come from the Valuer-General’s older annual editions; 2004 and 2005 were published under terms that reserved all rights, so they are not used.
5. Pooled. The average gap over the projects of a kind, with a 95% interval from 2,000 draws that resample projects and then suburbs within them. Fewer than four suburbs, or one project, gives no interval. A suburb near both a new station and a new centre is counted apart, as “near both”.
6. Decided first. Each kind is judged on one outcome named before the results: new stations, the gap two years after opening; new centres, three years after the announcement.

The projects

Each date links to its source. * Near both a new station and a new centre: counted apart. Caroline Springs station: it stands in Ravenhall, which has too few sales for a reliable median, and no other suburb’s centre is near, so there is no suburb to test. Cobblebank station: it stands in Cobblebank, which has too few sales for a reliable median, and no other suburb’s centre is near, so there is no suburb to test. Toolern Metropolitan Activity Centre: it stands in Cobblebank, which has too few sales for a reliable median, and no other suburb’s centre is near, so there is no suburb to test. Metro Tunnel: Shelved from 2011 to 2014 and recommitted in 2014; the route and stations were confirmed in 2015. Suburban Rail Loop East: Its six station precincts are being planned for more homes and jobs around the stations, so the line and the precincts cannot be told apart. Toolern Metropolitan Activity Centre: Planned around the station that opened as Cobblebank in 2019, so the centre and the station cannot be told apart after 2017. Westfield Doncaster expansion: Doncaster is also to have a Suburban Rail Loop North station, announced in August 2018. Arden precinct: Planned around the Metro Tunnel’s Arden station, so the precinct and the station cannot be told apart.

When the gap opens

The figures

The gap by years from the announcement

Years from the announcement	New stations: gap (95% interval), suburbs	New centres: gap (95% interval), suburbs
−9	—	−1.7% (−4.5% to +0.9%), 5
−8	—	−1.1% (−7.4% to +6.1%), 5
−7	—	+2.4% (−1.7% to +7.5%), 7
−6	—	+0.6% (−3.9% to +5.0%), 7
−5	—	+1.1% (−2.3% to +4.3%), 7
−4	—	−1.7% (−4.3% to +0.2%), 7
−3	+0.6% (−3.5% to +4.2%), 8	−2.9% (−7.4% to +1.2%), 10
−2	−2.4% (−10.1% to +2.4%), 8	+0.2% (−6.5% to +5.4%), 10
−1	0.0% (0.0% to 0.0%), 8	0.0% (0.0% to 0.0%), 10
0	−1.0% (−3.9% to +3.4%), 8	+0.4% (−4.8% to +6.4%), 10
+1	−6.6% (−11.4% to −2.8%), 8	−1.5% (−7.7% to +4.7%), 10
+2	−2.2% (−7.6% to +5.2%), 8	−1.7% (−5.9% to +3.4%), 10
+3	+2.8% (−4.2% to +12.6%), 8	−5.0% (−8.5% to −1.8%), 10
+4	−3.3% (−7.9% to +2.3%), 8	−4.5% (−10.9% to +2.9%), 10
+5	−5.0% (−10.0% to −1.6%), 7	−2.9% (−8.7% to +5.0%), 10
+6	−9.9% (−21.5% to +0.2%), 7	−3.6% (−12.8% to +8.1%), 10
+7	−0.6% (−10.9% to +10.0%), 7	−5.0% (−12.3% to +2.3%), 8
+8	+1.4% (−4.3% to +7.4%), 7	−4.8% (−11.1% to +1.3%), 8
+9	+0.7% (−5.8% to +8.8%), 7	−4.7% (−9.6% to +2.0%), 8
+10	−4.7% (−13.8% to +1.5%), 7	−1.9% (−6.8% to +4.3%), 8
+11	−7.1% (−9.7% to −4.3%), 4	−6.9% (−12.2% to −1.2%), 5
+12	−10.7% (−13.9% to −6.5%), 5	−4.7% (−11.1% to +2.7%), 5

Against its matched suburbs, a suburb that got a new station was −1.0% in the year it was announced and −1.2% in the year it opened (Valuer-General Victoria). Two years after opening, the outcome decided on first, the gap was −2.2% (95% interval −4.4% to +0.4%; 4 suburbs). For new centres, hubs and precincts, three years after the announcement, it was −5.0% (95% interval −8.5% to −1.8%; 10 suburbs). The only gaps clear of zero are falls: for stations at the route’s confirmation, −10.1%, with the Metro Tunnel’s the largest (−13.4%).

Were the trends parallel before? Matching uses the years just before the announcement, so only the years before those test it. For stations the gap there averaged +9.0% (3 suburbs, too few for an interval); for centres the gap there averaged −0.2% (95% interval −3.1% to +2.6%; 7 suburbs), consistent with parallel trends. For the projects announced in 2008 (Regional Rail Link, Caroline Springs station, Metro Tunnel, East Werribee Employment Precinct) the prices start too late to test it at all.

How near is near. The outcome decided on, with the centre’s distance set at 1 km and 3 km as well (a suburb a station or site stands in counts at every distance; Valuer-General Victoria):

Each project

One project’s gap rests on one to seven suburbs, so a single suburb’s year moves it; the pooled figures above are the ones to read (Valuer-General Victoria).

Tested on the newer projects

The older projects (Regional Rail Link, Mernda rail extension, Fishermans Bend urban renewal area, East Werribee Employment Precinct, Chadstone Shopping Centre expansion) give the prediction; the newer ones (East Pakenham station, Metro Tunnel, Suburban Rail Loop East, Footscray priority precinct and new hospital) test it, with any suburb near a newer project left out of the prediction. For stations the prediction missed by 5.0 points on average, against 6.7 for predicting no effect; for centres, 8.8 against 7.0 (Valuer-General Victoria).

What published studies found

• Yen, Mulley, Shearer and Burke (2018), “Announcement, construction or delivery: When does value uplift occur for residential properties? Evidence from the Gold Coast Light Rail system in Australia”, Land Use Policy 73, 412–422. A difference-in-differences study of the Gold Coast light rail: prices near stations began to rise after the announcement, rose most once the government committed the funding, and then slowed through construction and operation.
• Chen, Yazdani, Mojtahedi and Newton (2019), “The impact on neighbourhood residential property valuations of a newly proposed public transport project: The Sydney Northwest Metro case study”, Transportation Research Interdisciplinary Perspectives 3, 100070. Around Sydney Metro Northwest the effect was generally negative while the line was announced (2008–2012) and positive during construction (2013–2019), when prices were 0.037% higher for each 1% closer to a station.
• McIntosh, Trubka and Newman (2014), “Can value capture work in a car dependent city? Willingness to pay for transit access in Perth, Western Australia”, Transportation Research Part A 67, 320–339. Along Perth’s new fast rail corridors, land values within walking catchments of stations were up to 40% higher, from land value data on over 460,000 households.
• Mulley, Ma, Clifton, Yen and Burke (2016), “Residential property value impacts of proximity to transport infrastructure: An investigation of bus rapid transit and heavy rail networks in Brisbane, Australia”, Journal of Transport Geography 54, 41–52. In Brisbane the uplift was relatively low and varied across the city, and nearness to a busway station carried more of it than nearness to a train station.
• Abelson, Joyeux and Mahuteau (2013), “Modelling House Prices across Sydney”, Australian Economic Review 46, 269–285. Across 626 Sydney suburbs, access to the CBD and the coast raised median house prices, but access to a sub-centre had no separate effect.
• Jenner and Tulip (2020), “The Apartment Shortage”, RBA RDP 2020-04. In the paper’s regression of Sydney apartment prices (its Appendix D), nearness to train stations and large shopping centres was jointly insignificant, which the authors found surprising and put down to features those places share that are hard to control for.

The four rail studies use single properties placed by address, so they can see a station’s effect within a walk of it; a suburb’s median cannot, and blends homes near and far. Like the Gold Coast study, this one looks for the lift at each milestone; it finds no clear lift at announcement, route, construction or opening. The Sydney Metro Northwest study found the announcement years negative and construction positive; here the gap in the year the route was confirmed was −10.1% and in the year construction began +0.7%. Like the Sydney suburb study and the RBA’s apartment regressions, it finds no lift from a commercial centre that the data can see.

With and without the effect

Illustrative only. Not a valuation, not a prediction to rely on, and not advice.

The ranges of Part C for suburbs near projects announced but not built, with the yearly effect this study measured: for SRL East, the gap built up each year from construction to opening on the lines that opened, +0.9% a year (95% interval −0.2% to +1.8%); for a development not yet under construction, the gap built up each year from the announcement, +1.1% a year (95% interval −1.6% to +6.4%). Ranges like these held 79% of two-year outcomes in the backtest (Valuer-General Victoria).

Each suburb by rule: every SRL East station suburb (Monash station has no suburb of its name), then the most populous suburb near each commercial project not under construction by 2025; a suburb near both kinds has the commercial effect alone applied. Westfield Doncaster expansion: Doncaster is also to have a Suburban Rail Loop North station, announced in August 2018. Arden precinct: Planned around the Metro Tunnel’s Arden station, so the precinct and the station cannot be told apart.

High-speed rail and New South Wales

We found no high-speed rail station allotted to a Victorian suburb, so there is nothing to test; if one is allotted, the gap in the year a station is announced, −1.0% here, is the closest guide, and only as a scenario. New South Wales’s bulk sales data, which would bring Sydney Metro, the Western Sydney Airport line and Bradfield’s new city centre, is published under CC BY-NC-ND 4.0: not for commercial use, and nothing derived from it may be shared, so a company’s published medians cannot be built from it; a commercial licence is needed. It was not used.

What the event study cannot show

• Few projects and few suburbs: suburbs new stations stand in are big and few have enough sales, so each estimate rests on a handful.
• The rule for “near” was first a suburb’s centre within 1.5 km alone. A first run showed that left out the suburbs the outer stations stand in, so the suburb a station or site stands in was added before the results here; the outcomes decided on, the matching and the dates were fixed before any run.
• A suburb’s median mixes homes near and far from a station, and new estates change which houses sell: a growth-area suburb’s median can fall behind as cheaper new houses arrive, whatever a station does.
• The matched suburbs are far from every project here, not from every development Melbourne had; and inner-city suburbs have no matches as close to the CBD, since every suburb there is near the Metro Tunnel.
• Town centres planned with the growth-area stations, and the SRL East precincts planned around its stations, cannot be told apart from the stations themselves.

Part E. What else matters

On 4 October 2026 the founder asked for the factors that matter and were not on the original list. We wrote down 40 candidates, each with why it could matter and its open source, and the tests each had to pass, and committed them on 5 October 2026, before any of the new data was read against a price, arrears or insolvencies. Three could not be tested with open data and stay on the list. The data: the Census, population and building approvals from the ABS, recorded crime, Vicmap Planning, APRA, the RBA and AFSA.

How it was tested

1. One correction for everything. All 70 tests, of every kind, are one family, corrected by the Benjamini–Hochberg method at a 5% false discovery rate: 42 survive it, and 32 survive Holm’s stricter correction, reported beside it.
2. Price level. Each candidate is added alone to Part B’s model of the 2025 median. It passes if it survives the correction and it raises the share explained, and lowers the error, on councils held out (ten folds).
3. Growth. Each candidate is added to the starting price, measured at the start of the window, fitted on 2016–2021 and tested on 2021–2025. It passes if it survives the correction, keeps its sign when refitted on the later window, beats the starting price alone there, and beats it in at least three of four other pairs of three-year windows.
4. Early warning. The change in arrears from now to one to eight quarters ahead, on the indicator now. The lead is chosen on the fitting years alone, and its p-value multiplied by the eight leads searched. Housing arrears are fitted to December 2023 and tested from March 2024; insolvencies fitted to December 2016, tested from March 2017. It passes if it survives the correction, beats “no change” on the test years and keeps its sign there.
5. Regional. Personal insolvencies by SA3 in Victoria, per 10,000 residents and relative to the state, year by year: fitted to 2016, tested from 2017.
6. Robustness checks, which decide nothing: Holm’s correction; for price levels, the sign in every council fold and in Greater Melbourne alone; for growth, all four other pairs; for early warning, the pandemic quarters left out and the lead chosen by correlation instead.

Every factor, by its correlation with price

All 30 suburb factors from Parts B and E, each correlated on its own with the 2025 median (log), strongest first, with a 95% interval (ABS Census; Valuer-General Victoria). The piece prints the first ten. Left out because they are partly the price itself: Median rent, and the gross rental yield; Median mortgage repayment as a share of median income; Median rent as a share of median income; Growth in the suburb’s median over the two years before; Price relative to the council’s median.

The 40 candidates and how each fared

A candidate passed if any of its tests passed (ABS; APRA).

Price level

Ten of 22 candidates passed. Together, with the founder’s model, they explain 90% of the differences between suburbs’ 2025 medians on councils held out, up from 80%, and the typical miss falls from 18% to 12% (Valuer-General Victoria; ABS Census).

Together. Several measure overlapping things, so the passing factors were also fitted at once, with the founder’s model. Those that still carry weight, with |t| of 2 or more and a loss when left out, are the ones the piece names (Valuer-General Victoria).

Rents and repayments rise with prices, so median rent, rent relative to income and repayments relative to income partly restate the price: they are a cross-check on a valuation, not a reason for it. Repayments relative to income carries most weight of all, which says buyers in dear suburbs have stretched further for them. The three the piece names, with the founder’s model, explain 89% on councils held out. A standard deviation is the typical spread between suburbs: residents per square kilometre, a factor of 8.4; growth in the number of homes over the previous five years, 6.1 points a year; population growth over the previous five years, 3.1 points a year; share of residents aged 25 to 39, 7.7 points; median rent, a factor of 1.3; growth in median household income over the previous five years, 1.3 points a year; median mortgage repayment as a share of median income, 4.2 points; median rent as a share of median income, 3.8 points; share of land zoned for growth, 14.0 points; share of land prone to bushfire, 42.3 points (ABS Census; Vicmap Planning).

Growth

One of 24 passed. Most that survived the correction on 2016–2021 changed sign or failed on 2021–2025: which suburbs lead changes with the cycle (Valuer-General Victoria).

Share of households renting passed, narrowly: 0.5% closer than the starting price on 2021–2025, and in the other pairs +0.04%, +2.28%, +3.97%, −2.81% (Valuer-General Victoria). It is the only growth factor here, and too small to change the ranges.

Arrears and early warning

Four of 22 indicator tests passed, all for housing arrears; none passed for personal insolvencies. Housing arrears give 20 quarters to fit, so every lead rests on fewer than 20 pairs, and is marked so (APRA; AFSA; RBA).

The benchmark, the founder’s unemployment taken as its change over a year, is tested the same way and is not in the family: for housing arrears, a lead of 8 quarters, correlation −0.55, −44% against “no change”; for insolvencies, a lead of 8 quarters, correlation −0.28, +3% against “no change” (ABS via RBA table H5). Unemployment’s level moves with arrears (Part A); its change does not lead them.

Read the high-LVR result with care. Fewer new loans at 80% LVR or more went before rising arrears, 7 quarters later: high-LVR lending peaked in the 2020–21 boom and fell as rates rose. It marks credit tightening; it does not mean high-LVR loans are safer, and APRA’s data by band shows the opposite (APRA).

Early arrears are the plainest lead, and the least sure. A rise in loans 30 to 89 days behind went before more non-performing loans 2 quarters later, but it passed 1 of its 3 checks: with the pandemic quarters left out, too few remain to choose a lead (APRA).

Regional early warning

Personal insolvencies by SA3, 65 Victorian regions, 2010–2016 to fit (AFSA; ABS population). For personal loans, credit cards and small amount loans.

The illustrative ranges

The one growth factor that passed, share of households renting, was given to the ranges of Part C as each suburb’s gap from its market, fitted only on windows that ended by each origin. Over 5,607 two-year outcomes from origins 2017–2024, the miss went from 13.21% to 13.27% and the share inside the range from 76.4% to 76.4%. That is no better, so the ranges are unchanged. (Valuer-General Victoria)

What Part E cannot show

• One cycle. APRA’s housing data starts in 2019; every early-warning lead rests on one rise and fall of arrears, and could differ in the next. With so few quarters even Newey–West standard errors are rough, and so are the p-values built on them.
• Associated, not caused. A factor that passed goes with prices or arrears once everything else tested is held equal; something untested may drive both.
• Testing 70 things finds some by chance. The correction keeps the expected share of false finds to 5%: of 42, about 2 may be.
• SA2s, not suburbs. Growth in homes, people, rents and incomes, and approvals, are the SA2’s around a suburb.
• Look-ahead. Zones and overlays are as at October 2026, also for windows from 2016; crime for the window from 2016 is the year to June 2017, the earliest in the agency’s current tables.
• Regional insolvencies are all personal insolvencies: AFSA withholds the business split in about a third of Victoria’s regional quarters.
• Not tested, for want of open data: suburb sales counts, the variable-rate share of housing credit, fixed-rate expiries. The DFFH Rental Report and the Jobs and Skills Australia vacancy index did not answer a script and have no archived copy; Census rents and the ABS’s vacancies stand in.
• Settled after the first run, and said so in the code: which passing factors the piece names (the first rule compared gains in different units), and crediting the growth factor with its slope alone in the ranges test (the whole line missed by 13.70%). Before any test ran: approvals per resident taken as a log, and the planning overlays resampled at full detail.

How each figure was made

• Regression. Ordinary least squares, written out in full and tested on invented data. Standard errors: Newey–West for quarterly and yearly series, heteroskedasticity-robust for suburbs.
• Part E. The candidates, the tests and what counts as passing were committed before the data was read against an outcome (src/research/factors.js); the corrections and the lead search are checked on invented p-values and series whose answers are known.
• Out of sample. Arrears fitted to December 2023, tested from March 2024. Suburb prices tested on councils left out, 10 folds. Growth fitted to 2016–2021, tested on 2021–2025. Markets fitted to 1989–2010, tested on 2011–2025, and forecast in real time from each September since 2005 with what was known then.
• Suburbs. 797 localities have a Valuer-General median; 766 match an ABS suburb by name, and 31 that do not (names the ABS does not use, or uses twice) are left out. Medians the Valuer-General marks “^” are not used.
• Distances. Straight lines from the ABS’s point for each suburb to Flinders Street Station and to the nearest of 320 railway stations in the current timetable. For growth, stations opened later (12, from the Victorian Government’s announcements) are left out of the earlier window.
• Coast. Derived: a suburb whose boundary comes within 300 m of the shore of the ABS’s state boundary, with its centre within 3 km of it. 105 suburbs qualify.
• The example suburbs are, in order: inner, in Greater Melbourne, within 8 km of the CBD (Richmond); middle-ring, in Greater Melbourne, 10 to 20 km from the CBD (Reservoir); outer growth, in Greater Melbourne, more than 25 km from the CBD, population up at least 20% from 2016 to 2021 (Craigieburn); near new rail, nearest station opened between 2016 and 2021, within 3 km (South Morang); regional, outside Greater Melbourne (Mildura); coastal, on the coast, as the price model defines it, outside Greater Melbourne (Torquay).

What this cannot show

• Arrears are national and aggregate; the housing series covers one cycle, and APRA changed its definition of a non-performing loan in 2022. Loan-level data, which the RBA’s papers use, is not public.
• A suburb’s median is the middle sale of the year, so a change in which houses sold moves it as much as a change in prices. The Valuer-General revised 425 of 6,806 suburb medians between editions; the typical revision was 0.5%.
• Land size and shop prices are council averages, not suburbs’. Distances are straight lines, not travel times. Income is a 2021 snapshot.
• Housing supply is tested in Part E through each suburb’s SA2 and Vicmap Planning, and credit conditions only nationally, against arrears; neither is measured loan by loan.
• One state. A second (New South Wales) would test whether the results travel; its sales data is not licensed for a company’s published use (Part D).
• The forecasts assume the mortgage rate stays where it is. With half a point either way from January 2027, Melbourne’s 2027 midpoint moves from +8.7% to +9.9%.

Sources

1. Bergmann (2020), “The Determinants of Mortgage Defaults in Australia – Evidence for the Double-trigger Hypothesis”, RBA RDP 2020-03.
2. Read, Stewart and La Cava (2014), “Mortgage-related Financial Difficulties: Evidence from Australian Micro-level Data”, RBA RDP 2014-13.
3. Kearns, Major and Norman (2020), “How Risky is Australian Household Debt?”, RBA RDP 2020-05.
4. RBA statistical table Interest Rates and Yields — Money Market (F1.1): cash rate target. Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026. Data: f1.1-data.csv.
5. RBA statistical table Indicator Lending Rates (F5): standard variable housing rate, owner-occupier. Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026. Data: f5-data.csv.
6. RBA statistical table Housing Lending Rates (F6): owner-occupier variable rates, new and outstanding. Reserve Bank of Australia, from APRA data, CC BY 4.0, retrieved 4 October 2026. Data: f6-data.csv.
7. RBA statistical table Consumer Price Inflation (G1): Consumer Price Index. Australian Bureau of Statistics, via the Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026. Data: g1-data.csv.
8. RBA statistical table Labour Force (H5): unemployment rate. Australian Bureau of Statistics, via the Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026. Data: h5-data.csv.
9. Quarterly Authorised Deposit-taking Institution Property Exposures statistics, June 2026. Australian Prudential Regulation Authority, CC BY 4.0, retrieved 4 October 2026. Data: Quarterly authorised deposit-taking institution property exposures statistics June 2026.xlsx.
10. Quarterly Authorised Deposit-taking Institution Performance statistics, September 2004 to June 2026. Australian Prudential Regulation Authority, CC BY 4.0, retrieved 4 October 2026. Data: Quarterly authorised deposit-taking institution performance-September 2004 to June 2026 (1).xlsx.
11. Statement on Monetary Policy, August 2026 — Outlook, Table 3.1 Detailed Forecast Table. Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026.
12. Saunders and Tulip (2019), “A Model of the Australian Housing Market”, RBA RDP 2019-01.
13. Victorian Property Sales Report — Time Series, Year Summary 2025 (by municipality, 1995–2026). Valuer-General Victoria (Department of Transport and Planning), CC BY 4.0, retrieved 4 October 2026. Data: year-summary-2025.xlsx, from the Internet Archive’s copy.
14. Statement on Monetary Policy — Forecast Archive (forecasts published November 2018 to August 2026). Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026. Data: smp-forecast-archive.xlsx.
15. RBA statistical table Demand and Income (H2): real household disposable income. Australian Bureau of Statistics, via the Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026. Data: h2-data.csv.
16. National, state and territory population: Estimated Resident Population, Victoria, quarterly (ERP_Q). Australian Bureau of Statistics (Data API), CC BY 4.0, retrieved 4 October 2026. Data: 1.3.TOT.2.Q.
17. Victorian Property Sales Report — Median House by Suburb Time Series, 2015–2025. Valuer-General Victoria (Department of Transport and Planning), CC BY 4.0, retrieved 4 October 2026. Data: houses-by-suburb-2015-2025.xlsx, from the Internet Archive’s copy.
18. Census of Population and Housing 2021, General Community Profile DataPack, Suburbs and Localities, Victoria. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: 2021_GCP_SAL_for_VIC_short-header.zip.
19. Socio-Economic Indexes for Areas (SEIFA) 2021, Suburbs and Localities. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: Suburbs and Localities, Indexes, SEIFA 2021.xlsx.
20. Australian Statistical Geography Standard Edition 3 (2021), Suburbs and Localities, points. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: the query.
21. GTFS Schedule — Victorian public transport timetables and stops. Department of Transport and Planning, Victoria, CC BY 4.0, retrieved 4 October 2026. Data: gtfs.zip.
22. Victorian Property Sales Report — Median House by Suburb Time Series, 2014–2024. Valuer-General Victoria (Department of Transport and Planning), CC BY 4.0, retrieved 4 October 2026. Data: houses-by-suburb-2014-2024.xlsx, from the Internet Archive’s copy.
23. Australian Statistical Geography Standard Edition 3 (2021), Suburbs and Localities, generalised boundaries. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: the query.
24. SRL Planning Area Declaration Boundary (SRL East). Suburban Rail Loop Authority, CC BY 4.0, retrieved 4 October 2026. Data: the query.
25. Public Transport Lines and Stops — Public Transport Stops. Department of Transport and Planning, Victoria, CC BY 4.0, retrieved 4 October 2026. Data: public_transport_stops.geojson.
26. Victorian Property Sales Report — Median House by Suburb Time Series, 2013–2023. Valuer-General Victoria (Department of Transport and Planning), CC BY 4.0, retrieved 4 October 2026. Data: Houses-by-suburb-2013-2023.xlsx, from the Internet Archive’s copy.
27. Victorian Transport Plan to build a sustainable future. Premier of Victoria; the Internet Archive’s copy of the page first published at www.premier.vic.gov.au.
28. Proposed route of Regional Rail Link revealed. Premier of Victoria; the Internet Archive’s copy of the page first published at www.premier.vic.gov.au.
29. Regional Rail Link: Milestones. Regional Rail Link Authority; the Internet Archive’s copy of the page first published at www.regionalraillink.vic.gov.au.
30. The Victorian Transport Plan: Caroline Springs Train Station. Department of Transport, Victoria; the Internet Archive’s copy of the page first published at www4.transport.vic.gov.au.
31. The Victorian Transport Plan: Caroline Springs Station, frequently asked questions. Department of Transport, Victoria; the Internet Archive’s copy of the page first published at www4.transport.vic.gov.au.
32. New station for Caroline Springs. Public Transport Victoria; the Internet Archive’s copy of the page first published at www.ptv.vic.gov.au.
33. Melbourne’s Newest Station Ready For Passengers. Premier of Victoria.
34. Only Labor Will Extend the Rail Line to Mernda. Premier of Victoria; the Internet Archive’s copy of the page first published at www.premier.vic.gov.au.
35. Getting On With It: Mernda Rail Project Released To Market. Premier of Victoria.
36. Our Mernda Rail extension is no small thing.. Premier of Victoria.
37. Mernda Rail Extension Opens Today Six Months Early. Premier of Victoria.
38. Ballarat Line Upgrade Bringing Extra Benefits For Locals. Premier of Victoria.
39. Images Revealed For A New Station At Cobblebank. Premier of Victoria.
40. Victoria’s Newest Station Opens At Cobblebank. Premier of Victoria.
41. Two new stations and Pakenham level crossing removals fast-tracked. Level Crossing Removal Project; the Internet Archive’s copy of the page first published at levelcrossings.vic.gov.au.
42. A New Station For Melbourne’s Newest Suburb. Premier of Victoria.
43. More Level Crossings To Go In Pakenham. Premier of Victoria.
44. 80 Level Crossings Gone Plus Two New Stations In Pakenham. Premier of Victoria.
45. Swanston St The Smartest Option For Melbourne Metro. Premier of Victoria.
46. Demolition Begins For Metro Tunnel And Five New Stations. Premier of Victoria.
47. One Sleep Until The Metro Tunnel Opens. Premier of Victoria.
48. Underground Suburban Rail Loop To Connect Victoria. Premier of Victoria.
49. New Authority And Preferred Route For Suburban Rail Loop. Premier of Victoria.
50. Works Start Today On Victoria’s Suburban Rail Loop. Premier of Victoria.
51. Fishermans Bend Urban Renewal (fact sheet, April 2015). Department of Environment, Land, Water and Planning.
52. Fishermans Bend – an exciting new inner city suburb begins. Premier of Victoria; the Internet Archive’s copy of the page first published at www.premier.vic.gov.au.
53. Melbourne @ 5 million. Department of Planning and Community Development, Victoria; the Internet Archive’s copy of the page first published at www.dpcd.vic.gov.au.
54. East Werribee Employment Precinct. Victorian Planning Authority.
55. Chadstone expansion to create 7,000 new jobs. Premier of Victoria; the Internet Archive’s copy of the page first published at www.premier.vic.gov.au.
56. Chadstone Development To Create 2,200 Jobs. Premier of Victoria.
57. Toolern Precinct Structure Plan. Victorian Planning Authority.
58. Major Expansion Approved For Westfield Doncaster. Premier of Victoria.
59. New Footscray Hospital Project Gets Underway. Premier of Victoria.
60. Booming Footscray Our Newest Priority Precinct. Premier of Victoria.
61. Better Care And Jobs: Work Begins On Footscray Hospital. Premier of Victoria.
62. New Footscray Hospital Another Step Closer To Opening. Premier of Victoria.
63. Jobs And Affordable Homes On The Way To Arden. Premier of Victoria.
64. Arden Precinct. Victorian Planning Authority.
65. Yen, Mulley, Shearer and Burke (2018), “Announcement, construction or delivery: When does value uplift occur for residential properties? Evidence from the Gold Coast Light Rail system in Australia”, Land Use Policy 73, 412–422.
66. Chen, Yazdani, Mojtahedi and Newton (2019), “The impact on neighbourhood residential property valuations of a newly proposed public transport project: The Sydney Northwest Metro case study”, Transportation Research Interdisciplinary Perspectives 3, 100070.
67. McIntosh, Trubka and Newman (2014), “Can value capture work in a car dependent city? Willingness to pay for transit access in Perth, Western Australia”, Transportation Research Part A 67, 320–339.
68. Mulley, Ma, Clifton, Yen and Burke (2016), “Residential property value impacts of proximity to transport infrastructure: An investigation of bus rapid transit and heavy rail networks in Brisbane, Australia”, Journal of Transport Geography 54, 41–52.
69. Abelson, Joyeux and Mahuteau (2013), “Modelling House Prices across Sydney”, Australian Economic Review 46, 269–285.
70. Jenner and Tulip (2020), “The Apartment Shortage”, RBA RDP 2020-04.
71. Regional population: estimated resident population by SA2 and above (ASGS Edition 3), 2001 onwards (ABS_ANNUAL_ERP_ASGS2021). Australian Bureau of Statistics (Data API), CC BY 4.0, retrieved 4 October 2026. Data: ERP.SA2+SA3..A.
72. Building Approvals by SA2 and above, from July 2021 (BA_SA2): dwelling units, all sectors, total residential. Australian Bureau of Statistics (Data API), CC BY 4.0, retrieved 4 October 2026. Data: 1.9.TOT.100.SA2..M.
73. Crime Statistics Agency Data Tables — Criminal Incidents by local government area, postcode or suburb/town, years ending June 2017 to 2026. Crime Statistics Agency, Victoria, CC BY 4.0, retrieved 4 October 2026. Data: Data_Tables_LGA_Criminal_Incidents_Year_Ending_June_2026.xlsx.
74. Vicmap Planning — planning scheme zones, overlays and Bushfire Prone Areas (as at October 2026). Department of Transport and Planning, Victoria, CC BY 4.0, retrieved 4 October 2026. Data: FeatureServer.
75. RBA statistical table Housing Loan Payments (E13): interest charged and excess payments relative to household disposable income. Reserve Bank of Australia, from APRA and ABS data, CC BY 4.0, retrieved 4 October 2026. Data: e13-data.csv.
76. Quarterly Personal Insolvency Statistics — people entering a new personal insolvency, by state, type and business involvement, from September 2007. Australian Financial Security Authority, CC BY 2.5 AU, retrieved 4 October 2026. Data: quarterly_personal_insolvencies.csv.
77. Census of Population and Housing 2021, Time Series Profile DataPack (2011, 2016 and 2021 on 2021 boundaries), Statistical Areas Level 2, Victoria. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: 2021_TSP_SA2_for_VIC_short-header.zip.
78. Australian Statistical Geography Standard Edition 3 (2021), Statistical Areas Level 2, generalised boundaries. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: the query.
79. RBA statistical table Labour Costs and Productivity (H4): Wage Price Index, year-ended growth. Australian Bureau of Statistics, via the Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026. Data: h4-data.csv.
80. RBA statistical table Growth in Selected Financial Aggregates (D1): housing credit. Reserve Bank of Australia, CC BY 4.0, retrieved 4 October 2026. Data: d1-data.csv.
81. Consumer Price Index, Australia: rents, index, weighted average of eight capital cities, quarterly (CPI). Australian Bureau of Statistics (Data API), CC BY 4.0, retrieved 4 October 2026. Data: 1.115522.10.50.Q.
82. Total Value of Dwellings: mean price of residential dwellings, Australia, quarterly (RES_DWELL_ST). Australian Bureau of Statistics (Data API), CC BY 4.0, retrieved 4 October 2026. Data: 5.AUS.Q.
83. Quarterly Personal Insolvency Statistics — regional time series by Statistical Area Level 3, from September 2007. Australian Financial Security Authority, CC BY 2.5 AU, retrieved 4 October 2026. Data: regional_quarterly_time_series.csv.
84. Australian Statistical Geography Standard Edition 3 (2021), State boundary of Victoria, generalised. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: the query.
85. A Guide to Property Values — all statistics (by municipality, 1988–2019). Valuer-General Victoria, CC BY 4.0, retrieved 4 October 2026. Data: All-statistics.xls, from the Internet Archive’s copy.
86. Census of Population and Housing 2016, General Community Profile DataPack, State Suburbs, Victoria. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: 2016_GCP_SSC_for_VIC_short-header.zip.
87. Socio-Economic Indexes for Areas (SEIFA) 2016, State Suburbs (2033.0.55.001). Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: log.
88. Australian Statistical Geography Standard Edition 3 (2021), Local Government Areas of Victoria, generalised. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: the query.
89. Australian Statistical Geography Standard Edition 3 (2021), Greater Melbourne, generalised. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: the query.
90. Median House Price by Suburb, 2006–2016 (annual). Valuer-General Victoria (Department of Environment, Land, Water and Planning), CC BY 4.0, retrieved 4 October 2026. Data: suburb_house_2016.xls, from the Internet Archive’s copy.
91. Median House Price by Suburb, 2007–2017 (annual). Valuer-General Victoria (Department of Environment, Land, Water and Planning), CC BY 4.0, retrieved 4 October 2026. Data: suburb_house_2017.xls, from the Internet Archive’s copy.
92. Median House Price by Suburb, 2008–2018 (annual). Valuer-General Victoria (Department of Environment, Land, Water and Planning), CC BY 4.0, retrieved 4 October 2026. Data: Houses-by-suburb-2008-2018.xls, from the Internet Archive’s copy.
93. Median House Price by Suburb, 2009–2019 (annual). Valuer-General Victoria (Department of Environment, Land, Water and Planning), CC BY 4.0, retrieved 4 October 2026. Data: Suburb_House_final.xls, from the Internet Archive’s copy.
94. Median House Price by Suburb, 2010–2020 (annual). Valuer-General Victoria (Department of Environment, Land, Water and Planning), CC BY 4.0, retrieved 4 October 2026. Data: Suburb_Housev4_14May2021.xls, from the Internet Archive’s copy.
95. Median House Price by Suburb, 2011–2021 (annual). Valuer-General Victoria (Department of Environment, Land, Water and Planning), CC BY 4.0, retrieved 4 October 2026. Data: Suburb_HouseV2021A.xls, from the Internet Archive’s copy.
96. Vicmap Planning — Planning scheme zones: the Capital City Zone at Fishermans Bend (Port Phillip schedule 1, Melbourne schedule 4). Department of Transport and Planning, Victoria, CC BY 4.0, retrieved 4 October 2026. Data: the query.
97. VPA Precinct Boundaries — East Werribee Employment Precinct. Victorian Planning Authority, CC BY 4.0, retrieved 4 October 2026. Data: the query.
98. Australian Statistical Geography Standard 2016, Statistical Areas Level 2, generalised boundaries. Australian Bureau of Statistics, CC BY 4.0, retrieved 4 October 2026. Data: the query.
99. Regional population: estimated resident population by SA2 and above (ASGS 2016), 2001 to 2021 (ABS_ANNUAL_ERP_ASGS2016). Australian Bureau of Statistics (Data API), CC BY 4.0, retrieved 4 October 2026. Data: ERP.SA2+SA3..A.
100. Building Approvals by SA2 and above, July 2016 to June 2021 (BA_SA2_2016-21): dwelling units, all sectors, total residential. Australian Bureau of Statistics (Data API), CC BY 4.0, retrieved 4 October 2026. Data: 1.9.TOT.100.SA2..M.
101. Building Approvals by SA2 and above, July 2011 to June 2016 (BA_SA2_201116), on 2016 boundaries: dwelling units, all sectors, total residential. Australian Bureau of Statistics (Data API), CC BY 4.0, retrieved 4 October 2026. Data: 1.9.TOT.100.SA2..M.
102. Craigieburn welcomes metropolitan rail services. Premier of Victoria; the Internet Archive’s copy of the page first published at www.premier.vic.gov.au.
103. Works start at new Coolaroo station. Premier of Victoria; the Internet Archive’s copy of the page first published at www.premier.vic.gov.au.
104. First trains to South Morang mark historic day for rail. Premier of Victoria; the Internet Archive’s copy of the page first published at premier.vic.gov.au.
105. Community event to mark opening of Lynbrook station on Sunday 22 April. Premier of Victoria; the Internet Archive’s copy of the page first published at premier.vic.gov.au.
106. Cardinia Road Station powers up for full services. Premier of Victoria; the Internet Archive’s copy of the page first published at www.premier.vic.gov.au.
107. Williams Landing Project. Public Transport Victoria; the Internet Archive’s copy of the page first published at ptv.vic.gov.au.
108. Southland Station Set For A Festive Opening. Premier of Victoria.
109. New Station For Melbourne’s West As Part Of Airport Rail. Premier of Victoria.

Census, SEIFA, boundaries, CPI, unemployment, income and population: based on ABS data, CC BY 4.0. Sales statistics: © State of Victoria (Valuer-General Victoria), CC BY 4.0. Timetable: © State of Victoria (Department of Transport and Planning), CC BY 4.0. RBA material: © Reserve Bank of Australia, CC BY 4.0. APRA statistics: © Australian Prudential Regulation Authority, CC BY 4.0. Crime statistics: © Crime Statistics Agency, Victoria, CC BY 4.0. Vicmap Planning: © State of Victoria (Department of Transport and Planning), CC BY 4.0. Personal insolvency statistics: © Commonwealth of Australia (Australian Financial Security Authority), CC BY 2.5 AU. The Valuer-General’s site asks a script to pass a browser check, which we did not attempt; its files were taken from the Internet Archive’s copies, which the Victorian Government’s data portal links to.