This brief tests four beliefs held by a buyer who can pay cash for an Auckland home and does not want a mortgage. First, that Auckland property has fallen in real terms. Second, that the mainstream investment case is weaker than its reputation. Third, that younger generations are changing their relationship with ownership. Fourth, that an AI-driven global crisis in 2027 is a serious risk. The report also tests the alternative those beliefs imply: that holding cash is the safer choice. In plain terms, it asks whether buying a home without a mortgage is a good use of this buyer's money, and whether waiting in cash would be safer.
The report does not forecast prices, assign probabilities to a crisis or recommend that anyone buy, rent or hold. It separates what the evidence shows from what is interpretation, and it labels the stress scenarios as assumptions.
The decline is real, and it is larger than the headline. Before inflation is counted, the QV Auckland house price index fell 24.8 percent from its January 2022 peak to September 2026. After inflation is counted, the fall is 37.7 percent, measured from the real peak in December 2021. Over five years the real fall is 32.6 percent. The longer record is more mixed. Over 15 years Auckland housing returned 45.7 percent after inflation, and over the 19 years the index covers it returned 27.5 percent, or about 1.3 percent a year. The Global Financial Crisis produced a milder Auckland fall of 11.8 percent before inflation and 15.9 percent after it. The recent decline is therefore exceptional, while the long-run return is modest rather than absent.
The investment case depends heavily on borrowing. Annual rent before costs is 2.92 percent of the purchase price, which is the gross rental yield. The average home costs about nine times median household income, and a two-year mortgage rate is 5.89 percent. A buyer who borrows at 5.9 percent to buy an asset yielding 2.9 percent has an interest bill larger than the rent, before insurance, rates, maintenance and vacancy. The bank's net interest margin, which measures net interest income as a share of interest-earning assets, was 2.30 percent across the banking system. Agents earn commission on each sale. This supports the buyer's scepticism about borrowing to invest in property. A home bought without a mortgage is a different product: it provides a place to live, avoids rent and carries no risk of a margin call (a demand to repay debt when an asset falls in value). It should be judged on those terms.
Home ownership has changed across generations, but the evidence does not show that people want to own less. Auckland household ownership fell from 73.9 percent in 1986 to 59.5 percent in 2023. Among people aged 30 to 34, it fell from 43.6 percent in 2006 to 37.4 percent in 2023. The share of 20-year-olds still living at home as a dependent child rose from 52.3 percent in 2018 to 53.0 percent in 2023. Delayed household formation, rising rents and housing stress all appear alongside the fall in ownership. No representative survey shows a shift in preference away from owning, so the evidence points to constrained demand rather than rejection.
The AI bubble is a plausible risk, not an established fact. The four largest hyperscalers, the biggest cloud and AI companies, spent about US$357.5 billion on property and equipment in fiscal 2025, against US$23.0 billion in fiscal 2015. The top ten holdings of the S&P 500 tracker fund are 38.4 percent of the fund. The Shiller CAPE, a long-run measure of US share valuations, is about 41.1 on the latest reprint, below its 1999 peak of 44.2 and above its 2021 level of 34.5. AI issuers are 41 percent of new riskier US corporate bond issuance that was not refinancing, even though they are 1 percent of the riskier-bond index. But the Nasdaq is only 5.3 percent below its October 2026 peak, so no correction has occurred yet.
History is the key warning. New Zealand real house prices rose 8.7 percent through the dot-com bust but fell 15.7 percent in the credit-led Global Financial Crisis. In the stress matrix later in the report, renting and holding cash beats buying in all six crisis cells. Buying wins in the reference case where no unwind occurs. So the cash buyer is not choosing between safety and risk. They are betting on how likely a crisis is, and on how long they can wait.
Cash is insurance, not a plan. At current rates a one-year term deposit pays 4.02 percent, which after 33 percent tax and inflation is a fall of 0.98 percent a year. A six-month deposit is 1.30 percent negative in real terms after tax. Over 2000 to 2026 the average after-tax real return was 0.46 percent a year, and it was negative in 25.6 percent of months. In 1987, at the inflation peak, a 33 percent taxpayer still lost about 6 percent a year in real terms, even though deposit rates were high. The depositor compensation scheme covers up to $100,000 per depositor per bank, not every dollar. The right role for cash is money you can get at quickly and use if prices fall, not a permanent store of all your wealth.
The advice that follows is conditional, not clever. Buy the home you will live in for a decade if it passes the survival test and the real-return test. Do not buy it as an investment at a 2.9 percent yield. Keep 12 to 24 months of spending in insured cash. Do not use a mortgage to make a thin yield look attractive. Decide on observable triggers, not a forecast.
The decision behind this brief is concrete. A buyer in Auckland has enough money to buy a home without a mortgage and is trying to decide whether to do so. The buyer has seen real declines in value and doubts the decades-old claim that property is the safest investment available. The buyer also suspects that a global crisis rooted in the AI investment boom could arrive in 2027, and that holding cash might be safer. Each of those views contains a testable claim, and each can be partly right.
The report asks one question: what does the verifiable evidence say for a buyer who will not take a mortgage, deciding whether to buy an Auckland home in 2026 or 2027? It looks at the real return to housing, affordability and leverage, generational demand, policy risk, and how housing and cash hold up in an extreme financial crisis. It then sets out what decision rules follow from that evidence.
Five propositions are examined as competing claims, not conclusions. First, Auckland housing has fallen in real terms, but a fall from a peak is not proof that it is cheap. Second, the investment case is largely a leverage and tax case, so an owner without a mortgage earns a different return. Third, generational change in home ownership may reflect constraint rather than a change in preference. Fourth, an AI-capital unwind is a plausible global shock that could reach Auckland through credit, funding, migration and unemployment, but its timing cannot be forecast. Fifth, cash is defensive but not risk-free, and its real return depends on tax and inflation more than on the advertised interest rate.
The report is written for readers considering a similar decision, not only for the author. It explains each term on first use and reports uncertainty where the evidence is thin. It does not use a house-price forecast, a probabilistic crisis estimate or a personal recommendation. Section 4 sets out the data and method. Section 5 reports the findings by dimension. Section 6 turns the evidence into a decision framework, a stress matrix and a short set of rules. Section 7 lists what further research would change the picture.
Auckland housing in 2026 is the end point of a long boom and a sharp reversal. The QV Auckland index starts in January 2007. At that point, QV's average equalised value, its estimate of the value of a typical Auckland home, was about $514,000. It peaked at about $1.54 million in January 2022 and fell to about $1.16 million by September 2026. The boom was supported by low and falling interest rates, rapid credit growth, constrained land supply, strong migration and a tax system that allowed investors to deduct mortgage interest. The reversal followed the tightening cycle that began in 2021 and continued through 2023. It was amplified by the removal of interest deductibility in 2021 and softened by its partial restoration in 2024.
Prices are only one part of the market. Gross rental yield, which is annual rent before costs divided by price, fell from 3.58 percent in 2011 to 2.20 percent in 2021 and recovered only to 2.92 percent in 2026. Median weekly rent for newly lodged bonds in Auckland reached $650 in August 2026, up 10.2 percent from $590 five years earlier. Consumer prices rose 23.7 percent over the same period, so real rents fell about 11 percent. Inflation in rates, maintenance and insurance adds to the cost of ownership without adding to the rent.
Policy settings changed repeatedly over the same period. Loan-to-value (LVR) restrictions were introduced in 2013, removed in 2020 and reinstated in 2021. The bright-line test, which taxes gains on a property sold within a set period, was extended and then shortened to two years for sales from July 2024. Foreign buyers were restricted in 2018. Interest deductibility, which lets investors claim mortgage interest against rental income, was phased out from October 2021, restored to 80 percent from April 2024 and to 100 percent from April 2025. Debt-to-income (DTI) limits were activated in July 2024, and LVR settings were eased in December 2025. The depositor compensation scheme began on 1 July 2025. The practical effect is that the tax and credit settings of the 2015 to 2021 boom no longer exist in the same form, and the market has not returned to its old pattern.
The second background is the AI-capital cycle. The largest technology firms have increased capital expenditure, or capex, on data centres, chips and computing infrastructure at an extraordinary rate. The four largest hyperscalers, the companies that run the biggest cloud and AI data centres, spent about US$357.5 billion on property and equipment in fiscal 2025, against US$23.0 billion in fiscal 2015. The International Monetary Fund (IMF) projects about US$3.4 trillion of AI-related capital expenditure through 2029. The financing is shifting from internal cash flow toward credit. The Bank of England reports that five AI hyperscalers were 3 percent of the outstanding US investment-grade bond stock at the end of 2025 but more than 15 percent of investment-grade issuance in 2026 so far. AI issuers were 41 percent of non-refinance US high-yield bond issuance in 2026 so far, against 1 percent of the high-yield index. Private credit, which is lending outside the banks, is estimated to fund a rising share of AI investment, from 9 percent in 2024 to 34 percent in 2025, on one OECD estimate cited by the Bank of England.
New Zealand is not at the centre of that cycle, which is why the transmission question matters. New Zealand is a small open economy. Its banks fund themselves partly offshore: about 16.5 percent of total funding in November 2025. It has a large KiwiSaver pool of $138.8 billion and a Super Fund with 57 percent of assets in listed equities. Its migration cycle has already turned down from its 2023 peak. A global repricing could reach Auckland through wholesale funding costs, unemployment, migration, household confidence and the exchange rate. It might not reach housing at all if it remains an equity valuation event. The difference between the dot-com bust and the Global Financial Crisis is the analytical centre of the stress test.
The evidence base is built from primary and operational sources. Prices come from the QV House Price Index public API. QV measures an average equalised residential value for the Auckland region, monthly from January 2007 to September 2026, which is its estimate of the value of a typical home. The QV index comes from a state-owned valuation company and is not the REINZ median sale price. It is labelled non-official throughout and is never combined with REINZ values. For context, REINZ reported an Auckland median sale price of $1,100,000 in July 2022, and its September 2026 media release reports an Auckland median price down 2.6 percent on a year earlier. The median and QV's average equalised value are different measures and are not interchangeable. Consumer prices come from Stats NZ. Rents come from MBIE Tenancy Services bond data and the Stats NZ rent index. Income comes from the Household income and housing-cost statistics release (HILS) and, for one observation, from the 2023 Integrated Data Infrastructure estimate.
Credit data come from the Reserve Bank of New Zealand (RBNZ): mortgage rates, household debt and income, investor lending, housing loans, bad loans, the system interest margin, and bank capital and funding ratios. The series codes are listed in the methodology note. The RBNZ website returned HTTP 403 to this research environment, so the series were retrieved from Internet Archive captures of the exact official data files. Capture dates are recorded in the source ledger. Cash returns use RBNZ term deposit rates, Stats NZ CPI, Inland Revenue marginal tax rates and the depositor compensation scheme briefing. AI data come from US Securities and Exchange Commission company filings, official SPY holdings and Nasdaq index history. They are supplemented by Yale Shiller data and financial-stability publications from the IMF, the Bank for International Settlements (BIS), the Bank of England and the European Central Bank (ECB). Census data come from Stats NZ home-ownership tables and figure data packages for 2006, 2013, 2018 and 2023.
Real values are computed with the national CPI, base June 2017 quarter equals 1000. Monthly prices are deflated with the CPI for the quarter containing the month. Real returns are shown for four matched windows (5, 10, 15 and 19 years) and for the GFC episode, because a single preferred window would hide the effect of the start date. Housing returns are shown before and after an assumed five percent round-trip transaction cost, which is the cost of buying and later selling. Cash returns are shown after tax at 17.5, 30, 33 and 39 percent marginal rates. Percent changes are distinguished from percentage points, and levels from changes.
The stress test translates the request for a crisis ten times larger than 2008 into a set of bounded scenarios. A literal tenfold repeat of the GFC would imply a price fall larger than 100 percent, which is impossible. A tenfold repeat of the GFC's rise in unemployment would imply a rate in the mid-30s, beyond any modern experience. The report therefore uses three levels of severity: AI correction, AI credit bust and systemic AI unwind. It tests each one under two sets of conditions: a credit bust with falling inflation, and a funding squeeze with rising inflation. The settings are based on the RBNZ 2025 solvency stress test, the RBNZ 2020 very severe scenario, the 2024 reverse stress test and the IMF 2017 adverse scenario. Each cell states assumptions for global equities, AI capital expenditure, New Zealand unemployment, cumulative inflation, Auckland house prices, rents, deposit rates and deposit access. The assumptions are judgements, labelled as such, with no probabilities. The reference row is a no-unwind base case, not a forecast.
The owning-versus-renting comparison uses a break-even rate. Net shelter return equals gross rent avoided minus assumed ownership costs: rates, insurance and maintenance. Owning beats renting plus cash when real house-price growth exceeds the cash real return minus the net shelter return, plus the annualised transaction cost. The framework is sensitive to the cost assumptions. The full calculation is published in the housing user cost dataset and explained in section 6.
Every source has a stable identifier, an exact locator and an archived snapshot. A source is used only after both its content and metadata pass verification. Snapshots are hashed and recorded in the reference check. The complete methodology, codebook, limitations and replication instructions are in the methodology note linked below.
| Term | Plain meaning | Why it matters here |
|---|---|---|
| Nominal | A price or return before inflation is counted. | The 24.8 percent fall is nominal. |
| Real | A price or return after inflation is counted. | The 37.7 percent fall and the 1.3 percent long-run return are real. |
| Gross rental yield | Annual rent before costs divided by the purchase price. | 2.92 percent is below the mortgage rate. |
| Price-to-income ratio | The average home price divided by median annual household income. | About 9 in 2025. |
| Loan-to-value (LVR) | The share of a property's price that is borrowed. | LVR limits cap high-debt lending. |
| Debt-to-income (DTI) | Total debt divided by annual income. | DTI limits restrict high-debt borrowers. |
| Bright-line test | A tax on the gain from selling a residential property within a set period. | The period was shortened to two years in 2024. |
| Interest deductibility | Whether investors can claim mortgage interest against rental income for tax. | It was phased out from 2021 and restored by 2025. |
| Net interest margin | A bank's net interest income as a share of its interest-earning assets. | 2.30 percent across the banking system. |
| Shiller CAPE | A long-run measure of share prices against average earnings over ten years. | 41.1 is high but below the 1999 peak. |
| Round-trip transaction cost | The cost of buying and later selling a property, including fees and commission. | Assumed to be 5 percent in this report. |
| Term deposit | A bank deposit for a fixed period at a fixed interest rate. | 4.02 percent for one year. |
| Capex | Capital expenditure: spending on long-lasting assets such as data centres and equipment. | Big-four capex was US$357.5 billion in fiscal 2025. |
| Deposit taker | A licensed bank or other institution that accepts deposits. | The guarantee is per depositor per deposit taker. |
The QV index peaked at 2,996.5 in January 2022 and stood at 2,254.2 in September 2026, a nominal fall of 24.8 percent. After adjusting for consumer prices, the real peak is December 2021 and the real fall is 37.7 percent. Consumer prices rose 21.9 percent between December 2021 and September 2026. The different peak dates matter: a real decline should be measured from the real peak, not the nominal one.
The window chosen changes the answer. Over five years Auckland housing lost 32.6 percent in real terms, or about 7.6 percent a year. Over ten years it lost 22.2 percent real, or about 2.5 percent a year. Over fifteen years it gained 45.7 percent real, or about 2.5 percent a year. Over the 19 years the index covers it gained 27.5 percent real, or about 1.3 percent a year. Long-run studies of housing across countries find real returns that are positive but vary widely (Jordà et al., 2019). After an assumed five percent round-trip cost, the 19-year net real return is about 1.0 percent a year. The GFC episode produced a fall of 11.8 percent nominal and 15.9 percent real, which is much smaller than the current decline.
These numbers support the buyer's observation and limit it at the same time. Auckland housing has suffered a genuine real loss, and the recent past is not a story of resilience. But a 19-year real return of about 1.3 percent a year is not nothing, and the 15-year figure shows that the post-GFC decade was exceptionally strong. The relevant question is not whether property has fallen, but whether forward real growth will clear the break-even rate described in section 6.
QV Auckland House Price Index, January 2007 = 1000, monthly to September 2026. The real series is divided by the national CPI for the quarter containing each month (June 2017 quarter = 1000), so it starts at about 1,214 in January 2007. Sources: QV public HPI API; Stats NZ CPI.
The nominal index peaked at 2,996.5 in January 2022 and the real index peaked at 2,646.3 in December 2021. By September 2026 the nominal index was 24.8 percent below its peak and the real index was 37.7 percent below its real peak. The real series falls more because consumer prices rose while house prices fell.
Gross rental yield, which is annual rent before costs divided by price, was 3.58 percent in 2011, 2.37 percent in 2016, 2.20 percent in 2021 and 2.92 percent in 2026. The rise from the 2021 low reflects falling prices and rising rents, but a 2.92 percent gross yield is still thin. It ignores rates, insurance, maintenance, vacancy, transaction costs and tax. The price-to-income ratio, which is average value divided by median gross household income, was 10.70 in 2023 and 9.14 in 2025; using the 2025 income for 2026 gives 8.88, which is an overstatement if incomes grew. The 2025 income figure is provisional because Statistics NZ has flagged an error in the year-ended June 2025 release, with corrections due in February 2027.
On the cost side, national consumer price indexes show local authority rates and payments up 6.18 percent in the year to the September 2026 quarter, property maintenance services up 6.10 percent and dwelling insurance up 0.1 percent. Those are national indexes, not Auckland Council schedules, so they describe direction rather than the exact bill. The ownership costs used in the break-even calculation are assumptions and are published in full.
Left: gross rental yield, (median weekly rent for new bonds x 52) / average equalised value. Right: average equalised value / median gross household income. The three price-to-income observations use different income measures: administrative data (IDI) for 2023, and the Household income and housing-cost statistics release (HILS) for 2025 and 2026. Sources: QV; MBIE Tenancy Services; Stats NZ HILS; Auckland Council IDI estimate.
The gross yield reached its low of 2.20 percent in 2021 and recovered to 2.92 percent by September 2026. Price-to-income eased from 10.70 in 2023 to 9.14 in 2025, as values fell and the income measure changed. The two ratios use different income measures and are not directly comparable, but both are high. The 2026 ratio reuses the 2025 income and is therefore provisional.
Because the answer changes with the start date, the report measures real returns over four matched windows and adjusts for an assumed five percent round-trip transaction cost. Over five years Auckland housing returned minus 7.6 percent a year in real terms, or minus 8.5 percent after costs. Over ten years it returned minus 2.5 percent a year, or minus 3.0 percent after costs. Over fifteen years it returned plus 2.5 percent a year, or plus 2.2 percent after costs. Over nineteen years it returned plus 1.3 percent a year, or plus 1.0 percent after costs. Cash after tax averaged between minus 1.7 and plus 0.1 percent a year across the same windows. The table below reports the same values, and the published data file contains every input.
| Window | Housing, nominal total | Housing, real total | Housing, net real annualised | Cash, real after tax annualised |
|---|---|---|---|---|
| 5 years | minus 16.7 percent | minus 32.6 percent | minus 8.5 percent | minus 1.70 percent |
| 10 years | plus 7.9 percent | minus 22.2 percent | minus 3.0 percent | minus 0.90 percent |
| 15 years | plus 110.2 percent | plus 45.7 percent | plus 2.2 percent | minus 0.07 percent |
| 19 years | plus 108.6 percent | plus 27.5 percent | plus 1.0 percent | plus 0.06 percent |
Annualised real returns by holding window. Housing is shown gross and net of an assumed five percent round-trip transaction cost; cash is the window-specific mean of six-month term deposit returns after 33 percent tax and CPI. Sources: QV; Stats NZ CPI; RBNZ B3; author calculations.
The five-year window is dominated by the 2021 to 2026 downturn and shows a real loss of about 8.5 percent a year after costs. The fifteen and nineteen-year windows remain positive after costs, at about 2.2 and 1.0 percent a year. Cash after tax was negative in the five and ten-year windows and close to zero in the fifteen and nineteen-year windows.
A two-year standard mortgage rate was 5.89 percent and the floating standard rate was 6.26 percent at 31 August 2026, when the official cash rate was 2.50 percent. The cash rate rose to 2.75 percent on 2 September 2026. Against a 2.92 percent gross yield, the interest cost alone exceeds rental income by 2.97 percentage points before rates, insurance, maintenance and vacancy. The system net interest margin was 2.30 percent in the September 2025 quarter, within a range of 1.82 to 2.66 percent since 2000. The housing loan yield was 4.81 percent in June 2026, against a total deposit cost of 2.41 percent in November 2025. The two figures are measured in different months. The gross gap of about 2.4 percentage points is not the same as profit.
Household credit stress is currently low. The housing bad-loan (non-performing loan) ratio was 0.56 percent in November 2025, against a GFC peak of 1.2 percent in October 2009. The system tier 1 capital ratio was 15.1 percent in March 2026, the core funding ratio was 89.5 percent against a 75 percent minimum, and offshore funding was 16.5 percent of total funding. The investor share of new mortgage lending was 19.5 percent in August 2026, down from a peak of 34.8 percent in 2016, and the first-home-buyer share was 20.5 percent. That figure is the RBNZ measure of first-home buyers as a share of new mortgage lending. A separate industry buyer-classification measure puts first-home buyers at about 30 percent of Auckland property purchases in the first quarter of 2026, which uses a different denominator and is not an official statistic. Household debt to disposable income was 127 percent at its last official observation in June 2021, and the interest burden was 5.3 percent, well below the 14.0 percent peak in September 2008.
Two conclusions follow. First, the buyer's instinct about leverage has support. At current yields, a mortgage converts a low-yield asset into a bet on capital gains. The bank earns its margin whether or not the bet pays. Second, the claim that only the bank and the agent benefit is too strong. Lenders bear funding and credit risk, and owners in the 2011 to 2021 upswing captured large gains. Leverage amplifies gains as well as losses. What the evidence supports is narrower and more useful: for this buyer, a mortgage would add a negative-carry bet on top of an already uncertain capital-gains bet.
Auckland household ownership fell from 73.9 percent in 1986 to 63.8 percent in 2006, 61.5 percent in 2013, 59.4 percent in 2018 and 59.5 percent in 2023. National household ownership was 66.9 percent in 2006, 64.8 percent in 2013, 64.5 percent in 2018 and 66.0 percent in 2023, the first national increase since the early 1990s. Auckland's rate was essentially flat, rising from 59.4 percent in 2018 to 59.5 percent in 2023, far below the national increase.
Individual ownership by age shows the same pattern. Among people aged 30 to 34, ownership fell from 43.6 percent in 2006 to 37.4 percent in 2023. The 35 to 39 group fell from 56.6 to 49.2 percent, the 40 to 44 group from 64.8 to 56.8 percent, and the 25 to 29 group from 22.9 to 20.8 percent. The fall was steepest between 2006 and 2013, a seven-year gap because the 2011 Census was cancelled after the Canterbury earthquake.
The constraint evidence sits alongside the home-ownership data. The share of 20-year-olds still living at home as a dependent child rose from 52.3 percent in 2018 to 53.0 percent in 2023. At age 25 it rose from 24.3 to 25.6 percent. Families with adult children increased 27.3 percent between 2013 and 2023 against 13.9 percent for all families. Household crowding rose from 5.0 percent of households in 2013 to 6.2 percent in 2023. Median weekly rent in Auckland rose from 450 dollars in 2018 to 540 dollars in 2023. None of these is a measure of preference, and no representative survey of the desire to own versus rent was verified. The evidence supports a story of delayed household formation and affordability constraint, not a generation that has abandoned ownership.
Left: individual home ownership rates for selected age bands, New Zealand, 2006, 2013, 2018 and 2023. Right: household ownership rate, Auckland region, 1986 to 2023. Sources: Stats NZ Census home-ownership tables and 2023 Census figure data.
Every prime-age band owns less in 2023 than in 2006. The 30 to 34 band fell from 43.6 to 37.4 percent, the 35 to 39 band from 56.6 to 49.2 percent and the 40 to 44 band from 64.8 to 56.8 percent. Auckland household ownership fell 14.4 percentage points between 1986 and 2023, with most of the fall before 2013.
The policy timeline is not a single regime. Loan-to-value limits were introduced in 2013 to slow high-leverage lending and were removed in 2020 during the pandemic. They were reinstated in 2021, tightened for investors, and progressively eased in 2023 and again in December 2025. The bright-line test was extended, then shortened to two years for sales from July 2024. Interest deductibility was removed from October 2021, restored to 80 percent from April 2024 and to 100 percent from April 2025. Debt-to-income limits were activated in July 2024. The first-home grant was discontinued in 2024. Deposit protection began in July 2025. Foreign buyers were restricted in 2018.
| Date | Setting | What changed |
|---|---|---|
| October 2013 | Loan-to-value (LVR) limits introduced | Lending above 80 percent LVR (deposit below 20 percent) capped at 10 percent of new lending. |
| August 2018 | Foreign-buyer legislation | The Overseas Investment Amendment Act was passed. Residential land became sensitive land for most overseas persons, with the main restrictions commencing later in 2018. |
| March 2021 | LVR reinstated | Owner-occupier limit 20 percent of new lending above 80 percent LVR; investor limit 5 percent above 70 percent. |
| October 2021 | Interest deductibility phased out | Residential investment property interest limitation began. |
| April 2024 | Interest deductibility restored to 80 percent | Partial restoration for residential investment property. |
| July 2024 | Debt-to-income (DTI) limits and two-year bright-line test | DTI limits activated with an LVR easing; the bright-line period was cut to two years. |
| April 2025 | Interest deductibility restored to 100 percent | Full deduction of interest on residential investment property. |
| July 2025 | Depositor compensation scheme | Cover of up to $100,000 per depositor per deposit taker, such as a bank. |
| December 2025 | LVR eased further | Owner-occupier limit raised to 25 percent above 80 percent LVR; investor limit eased. |
The table describes settings, not measured effects. The report does not claim that any single setting caused a given change in prices; interest rates, supply, migration and income growth moved at the same time. The useful conclusion for a buyer is that the regime has become less generous to leveraged investors and more focused on financial stability, while remaining supportive of owner-occupiers with stable income. That is a reason to evaluate a home as a home, not as a tax-efficient investment vehicle.
The evidence supports concern about the scale and financing of the AI boom, but not a claim that the boom is certainly a bubble. The four largest hyperscalers spent about 357.5 billion US dollars on property and equipment in fiscal 2025, against 23.0 billion in fiscal 2015. The six largest technology firms in the dataset had combined capital spending of 470.6 billion dollars in their latest reported fiscal years, and 593.7 billion across their latest four reported quarters. The top ten companies account for 38.4 percent of the S&P 500 tracker fund, and NVIDIA alone accounts for 7.9 percent. The Shiller CAPE, a long-run valuation measure, is about 41.1 on the latest reprint, compared with 43.8 in January 2000 and 34.5 in January 2021. The Nasdaq Composite is 26,424.8 and only 5.3 percent below its 6 October 2026 peak, so the correction has not happened.
The financing evidence is the more relevant warning. Five AI hyperscalers were 3 percent of the outstanding stock of safer US corporate bonds at the end of 2025, but more than 15 percent of new issuance of those bonds in 2026. AI issuers were 41 percent of new riskier US corporate bond issuance that was not refinancing, against 1 percent of the riskier-bond index. The OECD estimate cited by the Bank of England puts the private-credit share of AI investment at 9 percent in 2024 and 34 percent in 2025. The ECB reports that up to 30 percent of a 3 trillion dollar data-centre buildout could be funded by private credit. Those are exactly the channels through which an AI capex reversal could become a credit event rather than a share-price event.
New Zealand's exposure is indirect but real. About 16.5 percent of bank funding was offshore in November 2025 and the core funding ratio was 89.5 percent, so a global funding shock would reach mortgage pricing before it reached house prices. KiwiSaver held 138.8 billion dollars at March 2026, including 68 billion dollars in growth funds, and the Super Fund held 93.5 billion dollars with 57 percent in listed equities. Net migration was 20,500 in the year to August 2026, down from a peak of 133,700 in the year to October 2023. A global repricing that reduced migration and lifted bank funding costs would hit Auckland demand and supply conditions at the same time.
Top left: big-four hyperscaler capital expenditure, fiscal years, US dollars billion. Top right: Shiller CAPE at three dates. Bottom left: S&P 500 top-ten weight and NVIDIA weight, October 2026. Bottom right: AI-related credit shares, percent. Sources: US SEC company filings; official S&P 500 tracker fund (SPY) holdings; Yale Shiller data and a third-party reprint; Bank of England and IMF publications.
The capex series rises about fifteen-fold between fiscal 2015 and fiscal 2025. Valuation and concentration are high but below the 1999 CAPE peak. The credit panels show the financing shift that would matter in an unwind. AI borrowers are a small share of outstanding debt but a large share of new issuance. Private credit is funding a rising share of the buildout.
Using the latest available settings (August to October 2026), the official cash rate is 2.75 percent, the 90-day bank bill is 3.13 percent, the six-month term deposit rate is 3.53 percent and the one-year rate is 4.02 percent. Consumer price inflation is 3.7 percent a year. The six-month deposit therefore returns 0.18 percent negative in real terms before tax, and 0.77 to 1.51 percent negative after tax depending on the marginal rate. The one-year rate returns 0.30 percent positive before tax and 0.98 percent negative after 33 percent tax. A depositor on the top marginal tax rate of 39 percent faces a 1.51 percent real loss on the six-month rate.
The long-run picture is better but still thin. Over the 320 months from January 2000 to August 2026, the average real six-month deposit return was 1.93 percent before tax and 0.46 percent after 33 percent tax, or 0.20 percent at 39 percent. The after-tax real return was negative in 25.6 percent of months at 33 percent. In the high-inflation episode of 1987 to 1991, the six-month rate reached 18.00 percent in July 1987, yet at the inflation peak a 33 percent taxpayer still lost about 6 percent a year in real terms. In 2008 to 2011, at the CPI peak of 5.3 percent, the after-tax real return was minus 2.35 percent. Cash preserves purchasing power only when the nominal rate exceeds inflation after tax, and that is not the current position.
Deposit protection is real but capped. The New Zealand Depositor Compensation Scheme began on 1 July 2025 and covers up to $100,000 per depositor per deposit taker, such as a bank. It is not a blanket guarantee, and it does not protect a large cash balance held with one institution. Cash also carries reinvestment risk and exchange-rate risk for a reader whose spending is in another currency. A crisis can produce a funding freeze or delayed access, even where deposits are ultimately protected.
Compared with housing, cash has better liquidity, better crisis optionality in a deflationary bust, no transaction cost and no maintenance bill. It has worse inflation protection, no shelter, no imputed rent and a hard guarantee limit. In an inflationary crisis, cash loses real value while nominal house prices may hold or rise. In a deflationary credit bust, cash protects purchasing power while house prices fall. The regime matters more than the label.
The findings do not resolve into a single verdict. They resolve into a set of conditions under which buying, waiting and holding cash make sense. This section sets out the owning-versus-renting arithmetic, the generational interpretation, the AI-unwind stress matrix, the role of cash and a short set of decision rules.
The average equalised Auckland home is worth about 1.16 million dollars. At a 2.92 percent gross yield, the rent avoided is about 33,800 dollars a year. Assume ownership costs of 1.50 percent of value (rates 0.35 percent, insurance 0.30 percent and maintenance 0.85 percent, all assumptions within stated ranges). That is about 17,392 dollars a year, leaving a net shelter return of about 16,408 dollars, or 1.42 percent of value a year. The owner also bears capital risk, illiquidity and transaction costs. The renter pays rent from income and keeps the capital in cash, which at a 4.02 percent one-year deposit and 33 percent tax returns 2.69 percent nominal and minus 0.97 percent real at 3.7 percent inflation.
The comparison comes down to a break-even rate: owning beats renting plus cash when real house-price growth exceeds the cash real return minus the net shelter return, plus the annualised transaction cost. At the current after-tax cash return of about minus 0.97 percent, the break-even rate is about minus 2.4 percent a year before transaction costs. With an assumed five percent round trip over ten years it is about minus 1.9 percent a year, and over five years about minus 1.4 percent a year. Using the long-run cash return of plus 0.46 percent a year, it is about minus 1.0 percent a year before costs. The table below shows the calculation; every assumption is published in the housing user cost dataset.
| Item | Value | Basis |
|---|---|---|
| Average equalised Auckland value | 1,159,440 dollars | QV, September 2026 |
| Gross rent avoided | 33,800 dollars a year (2.92 percent) | MBIE bond median rent, 650 dollars a week |
| Total ownership costs | 17,392 dollars a year (1.50 percent) | Assumption: rates 0.35, insurance 0.30, maintenance 0.85 |
| Net shelter return | 16,408 dollars a year (1.42 percent) | Rent avoided minus costs |
| Cash after tax | 2.69 percent nominal; minus 0.97 percent real | 4.02 percent one-year deposit; 33 percent tax; 3.7 percent CPI |
| Owning break-even rate, current cash return | About minus 2.4 percent a year | Cash real return minus net shelter return |
| Owning break-even rate, 10-year hold with 5 percent round trip | About minus 1.9 percent a year | Break-even rate plus annualised transaction cost |
| Owning break-even rate, long-run cash return | About minus 1.0 percent a year | 0.46 percent long-run cash real return |
The interpretation is concrete. If Auckland real house prices fall more than about 1 to 2 percent a year over the buyer's horizon, renting and holding cash wins financially. If they fall less or rise, owning wins, before the non-financial value of security and control. Over the last five years real prices fell about 7.6 percent a year, so renting and cash clearly won. Over the last 19 years they rose about 1.3 percent a year, so owning won. The decision is not a referendum on whether property is a good investment in general. It is a judgement about two things. Will the next decade's real growth beat a break-even rate of roughly minus 1 to minus 2 percent a year? Can the buyer hold through a crisis without being forced to sell?
The view that a new generation has changed its mind about housing is partly supported and partly not. Ownership has fallen for younger cohorts, and household formation is delayed. But the same data show rising rents, crowding and adult children staying longer. They also show a first-home-buyer share that rose above 27 percent in early 2026 and about 30 percent in Auckland in the first quarter of 2026, on industry buyer-classification data rather than an official statistic. The average age of an Auckland first-home buyer was 37 in 2025, two to three years older than in 2019. That is a picture of people still trying to buy, later and with more help, not a generation that has rejected the idea.
A genuine preference shift would show up in different evidence: longer and more satisfied renting, lower stated desire to own, different household forms, and policy that treats secure renting as a long-term option. The available survey evidence is thinner than the claim requires. The Ipsos Issues Monitor in September 2026 found housing named among the three most important issues by 19 percent of respondents, fifth, behind inflation at 63 percent, healthcare at 44 percent, the economy at 34 percent and unemployment at 21 percent. That measures how often people name housing as an issue, not whether they want to own. The 2023 General Social Survey finding that 22 percent of people in non-owner households would prefer a more suitable home, against 10 percent of owners, measures housing suitability. Neither establishes a change in the desire to own.
The practical conclusion is that generational change is likely to be slow, so it is not a reason to expect a sudden fall in demand. It may reduce the future rate of capital gain, but it does not by itself make ownership undesirable, and it may reverse if prices and rents move in favour of buyers.
The request for a 2027 crisis ten times larger than 2008 cannot be answered with one number. A tenfold repeat of the Auckland GFC price fall of 11.8 percent would be a fall larger than 100 percent. A tenfold repeat of the GFC's rise in unemployment, about 3 percentage points, would put the rate in the mid-30s. The report therefore uses a bounded ladder anchored to the most severe published stress tests. The RBNZ 2025 solvency stress test assumed unemployment of 10.5 percent and house prices 35 percent lower. The RBNZ 2020 very severe scenario assumed unemployment of 17.7 percent and residential and commercial property prices 48 percent lower. The RBNZ 2024 reverse stress test had banks designing scenarios with unemployment of 8 to 18 percent and house prices 30 to 55 percent lower. The IMF 2017 adverse scenario assumed unemployment above 10 percent and house prices 40 percent lower. The equity assumptions in the matrix are calibrated to past technology sell-offs: the Nasdaq fell 77.9 percent in 2000-02, 55.6 percent in 2007-09, 30.1 percent in 2020 and 36.4 percent in 2021-22. The tail cell in this report sits near the most severe of these, not beyond the limits of the possible.
Six cells are modelled, plus a no-unwind reference. In the AI correction cells, global equities fall 25 percent and AI capex stops growing. Unemployment rises to 5.5 percent, Auckland nominal prices fall 5 percent in the disinflationary case or 3 percent in the inflationary case, and real prices fall 6.9 or 8.1 percent. In the AI credit bust cells, equities fall 40 percent, AI capex falls 25 percent, unemployment reaches 8.5 percent, and Auckland nominal prices fall 20 percent or 12 percent depending on the regime, with real falls of 20.8 or 17.8 percent. In the systemic unwind cells, equities fall 65 percent and AI capex falls 60 percent. Unemployment reaches 15 percent, and deposit access is impaired. Auckland nominal prices fall 45 percent in the deflationary case or 25 percent in the inflationary case, with real falls of 44.4 or 31.2 percent.
The verdict is consistent across the six crisis cells: renting and holding cash produces a better two-year real position than buying. The cash return in the tail cells assumes that protected deposits are ultimately accessible; a depositor above the $100,000 per depositor cap, or in a failed institution, would do worse. In the disinflationary tail, the owner loses about 41.6 percent of the initial position while the renter's cash gains about 7.4 percent over two years. In the inflationary tail, the owner loses about 28.4 percent while the renter loses about 3.0 percent. The no-unwind reference reverses the result: buying gains about 1.4 percent while renting and cash loses about 0.3 percent over two years. The matrix therefore shows what waiting in cash costs if no crisis arrives. It is not a guaranteed gain. Waiting in cash is a bet that the crisis arrives; buying is a bet that it does not, or that the buyer can hold through it.
The historical analogues explain why the AI label is not enough. New Zealand real house prices rose 8.7 percent through the dot-com bust from 2000 to 2002. That period was an equity valuation event rather than a credit event, and interest rates fell. New Zealand real house prices fell 15.7 percent in the GFC, which was a credit event with a funding freeze and a global recession. The current AI boom has both an equity valuation channel and a credit channel. The equity channel alone is not a housing disaster; the credit channel is. That is why the report watches AI issuance shares, private-credit exposure and bank funding costs rather than the Nasdaq alone. It is also why the inflation outcome matters. From 1970 to 1990, New Zealand real house prices rose 34.7 percent even though inflation peaked near 18.9 percent. Housing provided partial long-run protection against inflation, while cash lost real value when inflation ran above deposit rates.
For the buyer, the scenario is informative rather than predictive. If the buyer is confident that an AI-driven credit event will arrive in 2027, waiting in cash is financially superior in every modelled cell. If the buyer is not confident, the no-unwind reference and the 19-year real return argue for buying, provided the buyer can hold through the stress. The decision cannot be reduced to a forecast, but it can be made robust to both outcomes.
Cash is useful for a specific job. It is the only asset you can use immediately, at no transaction cost, and its dollar value does not fall when markets fall. In a deflationary credit bust it preserves purchasing power while house prices fall. It is also the natural source of money to use if prices fall and a buyer wants to enter. But cash is a poor long-run store of value at current rates. After tax and inflation it loses about 1 to 1.5 percent a year. It also lost about 6 percent a year in real terms at the 1987 inflation peak, when inflation ran above deposit rates. It carries bank risk above $100,000 per depositor per bank, and it offers no shelter, no rent saved and no protection against a housing shortage.
A sensible cash strategy therefore splits your money in two parts. One part is enough insured cash for 12 to 24 months of spending, and for buying if prices fall, spread below the guarantee cap where necessary. The other part is diversified rather than concentrated in one asset or one currency. Do not treat a term deposit as an investment strategy, and do not measure the safety of cash by its nominal balance. The relevant measure is real purchasing power after tax, and on that measure cash is currently negative.
The evidence supports a short set of rules for a buyer who does not want a mortgage. They are deliberately conditional, because the data do not support a universal buy or wait call.
Triggers that would support buying:
Triggers that would support waiting:
No single trigger is decisive, and the list should be adapted to the buyer's circumstances.
Two non-financial factors matter too. The first is security: an owner-occupied home cannot be ended by a landlord's decision. In the 2023 Census, people in owner-occupied homes had lived at the same address for 10.6 years on average, against 4.6 years for people in non-owner-occupied households. The second is the option to customise, renovate and stay. Those are real benefits that the break-even calculation does not monetise, and they can justify buying even when the financial advantage is marginal.
The conclusion is conditional on evidence that could change. Several outcomes would strengthen the case for buying:
Several outcomes would strengthen the case for waiting:
The report can be checked against these conditions.
Hypothetical scenarios, author assumptions anchored to RBNZ, IMF and BIS stress tests and historical episodes. Values are percentage changes over a two-year window; the owner position excludes transaction costs and assumes the owner holds through the period. No probabilities are assigned. The reference row is a base case, not a forecast. Sources: stress anchors and historical analogues in the evidence ledger.
In all six crisis cells, renting and holding cash outperforms buying over two years. In the disinflationary tail the owner loses about 41.6 percent while the renter gains about 7.4 percent; in the inflationary tail the owner loses about 28.4 percent while the renter loses about 3.0 percent. In the no-unwind reference, buying gains about 1.4 percent while renting and cash loses about 0.3 percent. The matrix measures the cost of insurance, not a certain outcome.
| Decision test | Question | Pass condition |
|---|---|---|
| Shelter or investment | Is this mainly a home to live in, or a return-seeking investment? | Pass as shelter if the home fits a ten-year living plan; fails as an investment if the return depends on capital gain. |
| Holding horizon | Can I hold without selling for at least ten years? | Pass at ten years or more with safe cash flows; fails if a sale within five years is likely. |
| Crisis survivability | Would a 35 to 50 percent price fall and 12 to 24 months in which you could not sell force a sale? | Pass if no, and reserves cover two years of costs; fails if the purchase relies on selling or refinancing. |
| Real return break-even test | Is real house-price growth likely to exceed about minus 1 to minus 2 percent a year? | Pass if expected real growth is above the break-even rate; fails if you expect a sustained real decline. |
| Cash safety | Is my cash below $100,000 per depositor per deposit taker (such as a bank) and enough for liquidity? | Pass if insured, diversified and sufficient; fails if all cash sits with one bank. |
| Liquidity | Do I keep 12 to 24 months of spending outside the home? | Pass if yes; fails if all capital goes into the property. |
| Alternatives | Have I compared the home decision with renting plus a diversified portfolio? | Pass if the alternative is explicitly costed; fails if cash versus house is treated as the only choice. |
| Leverage | If I used a mortgage, would the yield cover the interest? | Not applicable for a cash buyer; a 2.92 percent yield against a 5.89 percent mortgage does not cover interest before costs. |
Four gaps limit what this brief can settle. The first is price data. Continuous free REINZ Auckland history is paywalled, so the report relies on the QV index and cannot test a 20-year window. A licensed REINZ series, or agreement to publish a historical extract, would materially improve the analysis. The second is debt. The official household debt-to-disposable-income series stops in 2021, which prevents a current measure of debt service and forces the report to use the 2021 observation and the interest burden series instead. A consistent current series would clarify how much buffer households actually have.
The third gap is preference. No representative, publicly verified survey of the desire to own versus rent was found. A longitudinal survey of home ownership and attitudes, linked to the census, would let researchers separate preference from constraint instead of inferring one from the other. The fourth is AI-credit exposure. New Zealand bank and non-bank exposure to AI-related credit, data-centre lending and private credit is not publicly disclosed in a comparable form. Better disclosure would allow a direct mapping from a global AI unwind to New Zealand funding costs and mortgage pricing.
Three smaller opportunities would improve future versions of this work:
Each is feasible with better data access, and none requires a new model.
All compiled datasets, the methodology note and the Python replication script are available below. The script reads the saved CSV files, checks that the required columns are present, prints the statistics used in this report and summary, and regenerates all six PNG figures from the same values. It does not embed chart-data arrays.
| File | Description | Download |
|---|---|---|
README_methodology.txt | Methodology notes, sources, compilation methods and known limitations | TXT |
replicate.py | Python replication script that regenerates every chart and statistic in the report | PY |
auckland_price_index.csv | Monthly Auckland QV house price index, CPI-deflated real index and percent change from peak | CSV |
affordability_yield.csv | Gross rental yield and price-to-income series, including the raw rent and value inputs | CSV |
real_returns.csv | Nominal and real totals, annualised real returns, transaction-cost adjustment and window cash returns | CSV |
ownership_cohorts.csv | Census home ownership rates by age band and the Auckland household tenure series | CSV |
migration.csv | New Zealand migration context series used in the demographic and transmission analysis | CSV |
cash_returns.csv | After-tax real term deposit returns at four marginal tax rates | CSV |
credit_stress.csv | Joined quarterly credit indicators including mortgage rates, arrears and funding ratios | CSV |
policy_timeline.csv | Dated policy settings for the bright-line test, interest deductibility, LVR and DTI limits | CSV |
ai_capital_cycle.csv | Chart-ready AI capex, valuation, concentration and credit panels | CSV |
nz_transmission.csv | New Zealand exposure indicators for the AI-capital transmission channels | CSV |
stress_anchors.csv | Selected RBNZ, IMF, BIS and historical stress-test anchor values | CSV |
housing_user_cost.csv | The annual ownership cost and hurdle calculation | CSV |
scenario_assumptions.csv | Assumptions behind each cell of the AI-unwind stress matrix | CSV |
scenario_outcomes.csv | Two-year real outcome values for each stress scenario | CSV |
transmission_channels.csv | The labelled transmission map from a global AI unwind to Auckland housing | CSV |
decision_tests.csv | The buyer decision checklist and its pass conditions | CSV |
analysis_statistics.csv | Every headline statistic printed by the replication script and quoted in the report | CSV |
The public files reproduce the reported measures and assumptions. They do not include paywalled REINZ reports or copyrighted source documents. Source locators and permitted local verification records are documented in the methodology.
The references below list the main publications and datasets used in this brief. Exact page, table or file locators for every number are recorded in the evidence source ledger, and archived copies are kept for audit.
| Source | Use in this report | Access or observation date |
|---|---|---|
| QV House Price Index API | Auckland nominal and real price path | 22 October 2026 |
| Stats NZ | CPI, census home ownership, household income, rents, migration | 2023 to 2026 releases |
| MBIE Tenancy Services | Auckland bond rents | August 2026 |
| Reserve Bank of New Zealand | Mortgage rates, credit, capital, funding and stress tests | Archived captures to October 2026 |
| Inland Revenue | Marginal tax rates and property tax settings | 22 October 2026 |
| SEC XBRL company facts | Hyperscaler capital expenditure | Filings to October 2026 |
| Bank of England, ECB, IMF, BIS | AI financing, stress anchors and financial stability | 2025 to 2026 publications |
| FMA and New Zealand Super Fund | KiwiSaver and Super Fund exposure | March and June 2026 |
| Ipsos New Zealand; Cotality | Housing issue salience; first-home-buyer shares and buyer classification | September 2026; 2025-2026 |
| REINZ | Official cross-reference at two dates | July 2022 and September 2026 |
The complete source and claim ledgers contain exact locators, archived snapshots, hashes and limitations. They remain in the project evidence folder for audit and replication.