Property still works in every one. Sydney top-bracket investors are where the returns get thinnest.

What’s actually on the table

The 12 May 2026 budget is approaching. Treasury has confirmed it is modelling changes to both negative gearing and the 50 per cent CGT discount. The Senate Select Committee on the CGT Discount, chaired by Greens senator Nick McKim, handed down its majority report on 17 March 2026, finding the discount distorts investment and skews housing toward investors.

Two policy levers dominate the publicly reported Treasury modelling:

• CGT discount reduction. The most-cited landing point is a cut from 50 to 33 per cent. Alternatives at 30 per cent or full abolition sit at the more aggressive end of the proposal space.
• Negative gearing restriction. The most-cited proposal is a two-property cap, where rental losses on a third or subsequent property could no longer be deducted from other income (salary, dividends). A one-property cap, advocated by the Australian Council of Trade Unions (ACTU) since August 2025, is the more aggressive alternative. A separate variant, restricting negative gearing entirely to new builds, was the form Bill Shorten took to the 2019 election and remains live in Labor backbench discussion.

The major unresolved variable is grandfathering. Most Labor signalling points to grandfathering existing holdings. The Greens explicitly oppose it. One government source cited in February 2026 reporting suggested retrospective application could raise around $5 billion a year, against $1 to $2 billion under grandfathering.

We’ve modelled five scenarios that bracket the most likely Treasury combinations on 12 May.

Three further proposals are publicly canvassed but sit outside our model: the Greens’ phaseout-with-no-grandfathering position, Senator David Pocock’s CGT 25 per cent on new builds, and the McKell Institute’s three-tier CGT. We cover each in the “What we are not modelling” section below.

What this does to year-one cashflow

We model two investor profiles. The typical investor holds one established property at the city’s median value. The portfolio investor holds three identical properties at the city’s median, the test point where both cap variants (S2’s cap of two and S3’s cap of one) are visibly differentiable. Both profiles assume 80 per cent LVR, a 6.25 per cent investor mortgage rate, 25 per cent operating-cost ratio, and a 10-year hold. Around 70 per cent of investor purchases are of established stock, so this is the cohort the modelling is calibrated to.

For a typical 1-property investor on the 37 per cent marginal tax rate, here is year-one after-tax cashflow per property under each scenario.

For S0, S1, S2, and S3, the year-one cashflow numbers are identical. The CGT discount cut (S1) only matters at sale, not in year one. The two-property cap (S2) doesn’t bite at one property. The one-property ACTU cap (S3) also doesn’t bite at one property — it only quarantines NG on the second property and beyond. So at the cashflow level, the story is really S0 / S1 / S2 / S3 versus S4.

Sydney is the most exposed. Under S4, the same property that costs $21,776 a year to hold under current rules costs $34,565 a year. That’s an extra $12,789 in year one, before any indexation, or roughly $1,066 a month. Brisbane sits at the next-highest exposure: a year-one S4 hit of $10,428 per year (S0 -$17,756 to S4 -$28,184).

Darwin sits at the other end. Its 6.0 per cent gross yield means the property is close to neutral cashflow even before tax. The S4 hit is just $1,146 a year, around $96 a month.

When the cap actually bites

For investors holding multiple properties, the cap design does matter. We modelled a three-property portfolio at each city’s median to expose where S2 and S3 differentiate.

The pattern is the same across every city. S0 and S1 produce identical year-one cashflow because the CGT cut only matters at sale. S2 disallows the negative-gearing offset on the third property. S3 disallows it on properties two and three. S4 disallows it on every established property.

In a Sydney three-property portfolio, each scenario step quarantines one more property’s NG — costing the portfolio $12,789, the same hit a single Sydney property takes under S4. The chart shows total annual portfolio cashflow: -$65,328 under S0/S1, -$78,117 under S2, -$90,906 under S3, -$103,696 under S4. Each $12,789 step is one more property losing its NG offset.

The 9.5 per cent of investors holding three or more properties, around 452,700 properties exposed to a two-property cap per ATO 2022-23 figures, are where the cap design matters. For the other 90.5 per cent, S2 and S3 are notational, not material.

What it does to 10-year after-tax wealth

The annual cashflow tells you how the property feels in your bank account. The 10-year after-tax wealth tells you whether the property earns its place in your portfolio.

Reading across the typical-investor table, the S1, S2 and S3 columns are identical because the negative-gearing cap doesn’t bite at one property. The story is really S0 versus S1 versus S4. Under the most likely outcome (S1, the CGT cut alone), every city loses 7.5 to 14.3 per cent of its S0 wealth. Sydney loses the most ($31,961 over 10 years for a 37 per cent bracket investor); Darwin loses the least ($15,319). The IRR drop is a uniform 0.5 percentage points across cities. Real money, but well within the range that a portfolio absorbs.

The cap-design distinction only shows up for multi-property investors. The same scenarios applied to a three-property portfolio at each city’s median:

Now the cap-design ladder steps down at every column. Each tightening (S1 to S2 to S3 to S4) removes the same chunk of 10-year wealth, the present value of one property’s NG offset over the hold period. In Sydney that step is $111,619, identical at every cap level; in Brisbane it’s $89,316; in Darwin just $2,690, because the 6.0 per cent yield means NG is barely doing any work there to begin with. From S1 to S4, a Sydney portfolio investor loses three of those Sydney-sized steps stacked, $334,857 in projected 10-year wealth, a 64.1 per cent reduction against S0.

S4 is where the case gets thinner for the single-property investor too. The S4 column in the typical table tells the story: Sydney drops to $80,000 of 10-year wealth, a 64.1 per cent fall against S0. At the 47 per cent bracket the same property drops 79.7 per cent. The latter is the thinnest cell at the central modelling assumptions (4 per cent capital growth, 6.25 per cent mortgage rate); IRR thins to 1.62 per cent at the 37 per cent bracket and stays positive but barely above a savings account return.

The property still works in every modelled cell. Even Sydney 47 per cent bracket S4 ends with positive 10-year wealth at those central assumptions. The structural question is what happens when the regime shifts away from them.

Where property stops working

Those central assumptions are realistic but not the only plausible combination. Real-world conditions can sit anywhere across the parameter space.

We ran a sensitivity grid for the typical 1-property investor at the top 47 per cent marginal tax rate (the more exposed bracket) under S4 (the most aggressive scenario). The central grid covers six cells: mortgage rates of 5.75, 6.25, and 6.75 per cent, against capital growth of 4 and 6 per cent. Of those six cells, just one turns capital-destroying: Sydney 47 per cent bracket S4 at a 6.75 per cent mortgage and 4 per cent growth, where the investor loses $5,242 over 10 years. Brisbane stays positive across the entire central grid.

That’s the central-grid finding. There is a more adverse regime worth acknowledging. If capital growth sustains at 3 per cent rather than 4 per cent, a regime Sydney experienced from 2017 to 2024 on Cotality data, the picture deteriorates. Across an extended 80-cell stress grid that adds 3 per cent growth and 7 per cent mortgage cases, 20 cells turn negative under S4 at the 47 per cent bracket. Sustained sub-trend growth is the variable that breaks the case for property, not tax policy. The structural thesis still holds: the dominant force in Australian property is supply tightness pushing prices up. The risk to that thesis is what happens if that force eases.

The headline city ranking is robust. Sydney is the most-exposed city in every cell of the extended grid. Darwin is the least-exposed in every cell. Hobart is consistently second-least-exposed; Canberra third. The ranking holds regardless of which mortgage rate or growth combination you favour.

Where exposure concentrates within each city

Citywide medians flatten what is genuinely a within-city story. The investor exposure score we built combines three LGA-level signals from Cotality data: renter share, gross rental yield, and median dwelling value. High renter share proxies investor concentration. Low yield means the property is more dependent on negative-gearing offset to make the cashflow work. High prices mean more leverage, more interest deduction, and more nominal capital gain exposed to a CGT discount cut.

The score is z-standardised within each capital city, so it identifies the most and least exposed LGAs relative to that city’s distribution. The liquidity filter excludes any LGA with fewer than 20 sales in the trailing 12 months.

The pattern is consistent across cities. The most-exposed LGAs are inner-city and inner-suburban. They have the trifecta: high renter share, sub-3 per cent gross yields, and median dwelling values well above their city’s average. The least-exposed are outer-suburban or peri-urban, with lower renter share, higher yields, and lower prices.

Top 5 most-exposed LGAs in each capital (selected):

• Greater Sydney: Woollahra, City of Sydney, North Sydney, Waverley, Mosman.
• Greater Melbourne: Stonnington, City of Melbourne, Boroondara, Port Phillip, Yarra.
• Greater Brisbane: Brisbane (CBD), Moreton Bay, Redland, Logan, Ipswich.
• Greater Perth: Peppermint Grove, Cottesloe, Mosman Park, Nedlands, Subiaco.
• Greater Adelaide: Adelaide (CBD), Unley, Walkerville, Norwood Payneham and St Peters, Burnside.
• Greater Hobart: Hobart, Kingborough, Clarence.

These are also, broadly, the LGAs where multi-property investors concentrate, so S2, S3, and S4 all bite differentially in these LGAs in ways our single-property model can’t show. Investors holding a third or fourth property in Stonnington or Woollahra will see their negative-gearing offset on those incremental properties quarantined under S2, more aggressively under S3, and entirely under S4.

The least-exposed cohort, by contrast, is the “still investable” map under aggressive reform. In Sydney, that’s Blue Mountains, Hawkesbury, Wollondilly, Sutherland, Central Coast. In Melbourne, it’s Moorabool, Yarra Ranges, Hume, Macedon Ranges, Cardinia. In Perth, it’s Mundaring, Swan, Serpentine-Jarrahdale, Murray, Kalamunda. The pattern is the same in every city: outer suburbs, lower median prices, higher yields, lower renter concentration. They were never the markets driving NG and CGT-discount headlines, and they aren’t the markets where the post-reform pencil sharpens to a point.

A note on Canberra: the ACT is administered as a single LGA at the ASGS level. We can’t sub-segment exposure within Canberra at LGA granularity. At the citywide level, Canberra sits in the middle of the eight capitals on every exposure metric.

What it means for new supply

The industry coalition (Property Council, HIA, MBA, REIA) commissioned modelling from Qaive and Tulipwood Economics in March 2026 with a headline figure of minus 45,500 dwelling starts over five years if negative gearing is removed except for one current property per investor. That averages roughly 9,000 fewer starts per year, or about 4 per cent of the current annual run-rate.

Two facts to hold alongside that estimate.

First, the current pipeline state. The most recent ABS Building Activity data (December 2025 quarter, seasonally adjusted) shows 53,567 dwellings commenced and 43,536 completed nationally. Annualised, that’s 214,000 commencements against 174,000 completions, against a National Housing Accord target of 240,000 a year. The pipeline is already 27 per cent below where it needs to be. Investor demand feeds into how many starts get financed, since pre-sales unlock development funding. But the binding constraint on completions is workforce, materials, and approval timelines.

Second, the cleanest natural experiment for a tax-driven investor pullback is the 2014 to 2016 APRA episode.

In the four quarters from Q1 2015 to Q1 2016, investor share of new lending fell from 44.8 per cent to 34.0 per cent, a drop of 10.8 percentage points. Combined Capitals dwelling values still rose 6.5 per cent year-on-year. The cash rate was essentially flat at 2.0 to 2.5 per cent across the entire window, which rules out monetary policy as the mover. Supply tightness absorbed the demand shift. New dwelling commencements continued ramping.

If a regulator-driven 10.8 percentage-point investor share cut couldn’t dent prices in the mid-2010s, the tax-driven version we’re now contemplating is unlikely to dent them in a market with a deeper supply shortfall. The mechanism doesn’t change just because the policy lever does.

Three historical precedents

Three episodes get cited frequently in this debate. They are worth remembering as the data actually records them.

1985 to 1987: The NG quarantine. Negative gearing was quarantined for two years. The industry standard interpretation is that rents rose, especially in Sydney. The data says rents rose around 1.3 per cent a year above wage growth in Sydney over the period; rents were essentially flat in Melbourne, Brisbane, Perth, and Adelaide. A localised effect, presented as universal.

1999: CGT discount introduction. The 50 per cent discount was introduced. Investor share of lending rose sharply. National prices accelerated through the early 2000s. The industry interpretation: the discount drives demand. The pro-reform interpretation: the discount juiced investor demand specifically, displacing first-home buyers. Both are evidence-based.

2014 to 2018: APRA macroprudential. Two waves: a 10 per cent investor lending growth cap (December 2014), then interest-only restrictions (March 2017). Investor share fell from 44.8 per cent peak to 28.5 per cent floor. Prices fell in 2018, but that overlapped with the royal commission, so causality is contaminated. The cleaner sub-episode is the 2014 to 2016 window, which is what we leaned on in the modelling section.

The pattern across all three episodes: investor demand responds to policy levers fast. Supply doesn’t. Prices reflect the gap.

What we are not modelling, and why

Three publicly canvassed proposals sit outside the modelled set above. We’ve sketched them here so the reader can place them in context.

The Greens position. The Australian Greens advocate restricting negative gearing and the 50 per cent CGT discount to a single investment property per investor, applied to all investors including existing portfolios. An investor currently holding multiple properties would keep both concessions on one property of their choice and lose them on every other property they own. Senator Nick McKim chaired the Senate Select Committee on the CGT Discount, which reported on 17 March 2026 finding the current discount distorts investment and skews housing toward investors. We didn’t model the full Greens position because Treasury isn’t costing it for the 12 May budget. The NG-cap component overlaps with our S3 scenario; the broader CGT restriction sits outside our modelled set.

Senator David Pocock’s proposal. Pocock advocates removing the CGT discount for residential properties purchased after 1 July 2026 (with grandfathering for pre-existing holdings), keeping a 25 per cent discount only for newly built homes held more than three years, plus a one-property NG cap on second and subsequent investment properties. The negative-gearing component overlaps with our S3. The CGT component differentiates new builds from established property at the discount level, and our model is scoped to investors buying established stock (around 70 per cent of investor purchases), so we didn’t run the new-build CGT differential. For an established-property buyer under Pocock, the result is functionally S3 plus full removal of the CGT discount on any property purchased after 1 July 2026.

The McKell Institute three-tier proposal. Holden and Cavanough’s July 2025 paper “Harnessing Aspiration,” restated in their February 2026 follow-up “Funding Fairer Housing,” proposes three different CGT discount rates: 70 per cent for new attached dwellings (apartments, townhouses), 50 per cent retained on new detached, and 35 per cent on existing detached. None of the tiers go to zero. The structural intent is to shift investor capital from established to new build, particularly attached, by widening the tax-incentive gap. We didn’t model this because it isn’t currently in publicly reported Treasury modelling. If it lands on 12 May or in subsequent reform, we’ll model it.

In every case, we’ll publish a post-budget update with whatever scenario actually lands.

What to watch before the budget

• The shape of the proposal between now and 12 May. Specific landing points (33 per cent vs 30 per cent CGT; two-property vs one-property cap vs new-builds-only NG) usually firm in the days before the budget reading.
• Any signal on grandfathering. This is the single biggest variable for existing portfolios. Full grandfathering means S1 outcomes for current holdings; partial or no grandfathering shifts everything toward S4 economics.
• Coalition response posture. If the Coalition opposes every element of the package without offering compromises, it preserves a clean policy-reversal pitch heading into the 2028 election.

Keep reading

• 
  How the 2026-27 Budget’s negative gearing reform affects your borrowing capacityMay 25, 2026
• 
  The first look at the market after the budget: capital city values just tipped negative. Regionals are still rising.May 20, 2026
• 
  The 2026 budget rewrote property tax. Finding the right asset is more crucial than ever.May 13, 2026