Equity investing asks what an asset might become. Lending asks what it is worth on a bad day, to a real buyer, inside 90 days. These papers are about the second question. Monthly data on the Sydney housing market, and analysis of duration, concentration and collateral for wholesale investors and their advisers.
Sydney dwelling values are falling. As at the June index, values were down 1.2% for the month, 3.2% for the quarter, and roughly flat over the year, sitting about 3.7% below their recent peak, with sales volumes down 17% on a year earlier.1 We publish that in the first paragraph of our first note because a lender who needs prices to rise has built the wrong book. Our loans are written at 65 to 70 cents against each dollar of independently valued security, which means the borrower's equity, and not our investors' capital, stands first in line for a market like this one.
We would also note what the downturn has not changed. Housing credit continues to grow, non-bank lenders continue to take share, and the Reserve Bank has begun asking harder questions about lending standards in private credit.2 We think those questions are fair, and Paper 06 is our attempt to answer them in public.
Starting valuation against the following decade, stylised. Cream points: S&P 500, starting Shiller CAPE (x axis) against subsequent 10 year annualised total return (y axis), monthly since 1881. Gold points: US high yield, same construction from index inception. Gold line: current CAPE near 42, the second highest reading on record.11 Point clouds shown are a stylised representation of the published relationship and will be regenerated directly from the Shiller and ICE BofA datasets, with code included, before publication.
The Shiller CAPE ratio closed July above 42 for the first time since July 2000. Nobody knows what equities return from here. That is not a figure of speech; the historical record from this starting point is a fan of outcomes centred near zero. Credit's fan has a different shape, for a structural reason worth understanding.
An equity return over a decade decomposes into three parts: the dividends collected, the growth in earnings, and the change in the multiple the market pays for those earnings. The first two are earned. The third is awarded, and at a CAPE of 42 the awarding has largely already happened. For the multiple merely to stand still, the market of 2036 must be willing to pay a near record price for earnings. History records that happening rarely, which is why the left side of our chart is crowded and the right side is sparse, low and wide.
A credit return decomposes into two parts: the yield purchased, minus the losses taken. There is no multiple term. This is why the gold points sit in a band while the cream points fan out, and it is also why the comparison should not be oversold: credit obeys the same law through its own variable, since a tight starting spread predicts thin credit returns exactly as a high starting CAPE predicts thin equity returns. The discipline in credit is refusing to lend when the price of risk is wrong, which is a decision made loan by loan rather than assigned by an index.
The paper closes at the limiting case. A first registered mortgage held to term has no traded price and no multiple, so the purchased yield and the realised return are the same number unless the borrower defaults, and the 35 cents of borrower equity in front of the loan exists for precisely that event. The entire investment question collapses into underwriting, which is where we think it belongs.
| Equity: dividends + earnings growth + re-rating | three terms, one unknowable |
| Traded credit: purchased yield − losses ± spread moves | the middle term is underwritable |
| Mortgage held to term: purchased yield − losses | two terms, both underwritable |
| At CAPE 42, the re-rating term must hold a record level | history is unkind to this |
Eight series, one page, every month. The same numbers our credit committee reads, published with their source identifiers so they can be checked. Where a reading is unfavourable to the case for property lending, it is published anyway; that is rather the point of the exercise.
| Series | Latest | Quarter | Annual | Source |
|---|---|---|---|---|
| Greater Sydney median dwelling value | $1,265,608 | −3.2% | +0.3% | Cotality HVI1 |
| Sydney monthly change in dwelling values | −1.2% | falling | Cotality HVI | |
| Decline from cyclical peak | −3.7% | Cotality HVI | ||
| Estimated sales volumes, Sydney | −17.0% yr | Cotality | ||
| National dwelling values, monthly | −0.7% | Cotality HVI, Jul7 | ||
| Auction share of new listings, national | ~30% | from 45% Nov 25 | Cotality Chart Pack8 | |
| Non-bank share of financial system assets | ~6% | rising | RBA FSR2 | |
| Loans 90+ days past due, ADI housing | TBA at launch | APRA QADIP |
Readings above are drawn from the June 2026 Cotality Home Value Index (data to 31 May and 30 June as noted), the July 2026 Cotality Housing Chart Pack, and the RBA Financial Stability Review of March 2026. Figures are re-verified against primary releases before each monthly publication; where a series has been revised by the provider, the revision is noted here rather than silently applied.
Everything we publish answers one of three questions. If a proposed paper does not, we do not write it.
Concentration, correlation and the gap between a portfolio's label and its factor exposure. A diversified index can still be one trade.
Duration, drawdown and the two different risks bundled into fixed income. Some defensive assets defend against the wrong enemy.
Collateral, seniority and recovery mechanics in New South Wales. The question a lender must answer before writing the loan, not after.
The Austrian century bonds as a controlled experiment in duration. AA+ credit, uninterrupted coupons, and a two-thirds capital loss, alongside the AusBond Composite's record 12.7% drawdown in 2022. Separates repayment risk from re-pricing risk, and shows which one a held-to-term mortgage actually carries.
What this paper does not claim: that rates will rise again, or that traded bonds have no place in a portfolio.
The Magnificent Seven now weigh roughly a third of the S&P 500, against about 13% in 2018, with the top ten near 40%, the highest concentration since the Nifty Fifty era.9 This paper measures what that does to an Australian balanced portfolio through the international equities sleeve, works through the June 2026 episode in which the cohort shed roughly US$2 trillion in days,10 and notes, in fairness, that the Seven underperformed the broader index across the first half of 2026. The argument is about measurement, not prophecy: an investor should know how much of their diversification is nominal.
What this paper does not claim: that a correction is imminent, or that private credit is uncorrelated with the economy. A recession reaches everyone; seniority and security determine in what order.
Before 2021, the AusBond Composite had recorded two negative years in three decades. Then came its worst drawdown on record, at the same moment as an equity bear market, which is precisely when the defensive allocation was supposed to earn its keep.5 The paper decomposes the index by duration and issuer, prices a 100 basis point move against it, and asks how much of the average defensive allocation is a levered position on the direction of long rates that its owner never chose.
What this paper does not claim: that 2022 will repeat. Duration cuts both ways, and at today's yields the cushion is real. The point is that the risk should be held knowingly.
A stress test rather than a forecast, made timely by the fact that the stress has started: Sydney is about 3.7% off its peak as we write. We take a first registered mortgage written at 65% LVR through the drawdowns of 1989 to 1991, 2017 to 2019, and 2022, then through a hypothetical 30% fall worse than any of them, and trace the loss waterfall in each case, including selling costs, interest arrears and time to realisation. The result is not that losses are impossible. It is that the first 30 to 35 cents of every fall belongs to someone else.
What this paper does not claim: that capital is protected, or that a severe enough correction, paired with borrower default, cannot produce a loss.
The unglamorous legal mechanics that decide recovery outcomes in New South Wales: indefeasibility under Torrens title, priority between registered and unregistered interests, the practical worth of a caveat, and mortgagee sale under the Real Property Act. Written for the investor who has been offered 4% more by a fund that holds second mortgages, and would like to know what the extra 4% is actually paying for.
What this paper does not claim: that second-ranking credit is illegitimate. It is a different risk at a different price; our view is only that the two should never be pooled without saying so.
The March 2026 Financial Stability Review notes strong growth in non-bank and private credit, some loosening of covenants and presale requirements in property lending, and, pointedly, that regulators have limited visibility into private credit standards.2 This paper takes the criticisms seriously instead of around. It sets out the specific disclosures we think any pooled mortgage fund should publish, including weighted average LVR, arrears, related party exposures and valuation policy, and commits Oxford to publishing its own against that list, on this page, from first deployment.
What this paper does not claim: that regulation is unnecessary, or that our sector's growth contains no risk. A fund that cannot survive scrutiny of its lending standards is not being harmed by the scrutiny.
Recovery is a function of buyer depth, not postcode prestige, and buyer depth is measurable. Using turnover, days on market and clearance rates across Greater Sydney's SA3 regions, the paper maps where security property actually transacts inside 90 days under current conditions, with sales volumes down 17% year on year. The output doubles as a disclosure of our own geographic lending criteria, which we would rather publish than imply.
What this paper does not claim: that any suburb is immune to illiquidity. In 1990 there were postcodes where nothing sold at any price for months; the paper includes them.
Equities and credit at today's starting prices. The Shiller CAPE ratio sits near 42, its second highest reading in roughly 150 years of data, exceeded only at the dot-com peak of 44.2 and above 42 for the first time since July 2000.11 Plotting starting CAPE against subsequent 10 year equity returns since 1881 shows the familiar downward fan: from readings above 35, realised outcomes cluster near zero to 3% a year, with wide dispersion. Credit occupies a different geometry. Its forward return is anchored to a contractual coupon and repayment at par, so the fan is narrower and centred on the yield actually purchased. The paper rebuilds this analysis from Shiller's public dataset and ICE BofA high yield index data, then extends it to the instrument we know best: a mortgage held to term, where the purchased yield and the realised return are the same number unless the borrower defaults.
What this paper does not claim: that equities should be sold, or that CAPE times anything. It first crossed its 1929 level in 1996, four years early. Credit has its own version of the same law, since a tight starting spread predicts thin credit returns exactly as a high starting multiple predicts thin equity returns. The honest statement is symmetrical: in both markets, the price paid at entry is most of the return story, and only one of the two writes that return into a contract.
The claim printed on nearly every private credit deck, audited from the inside. Measured correlation between mortgage funds and equities is close to zero, and the paper begins by explaining the unflattering half of why: a loan book with no traded price cannot register volatility, so part of the smoothness in the sector's return charts is an artefact of not looking. Critics call this volatility laundering, and they are partly right. The paper then sets out what survives the critique, which is the part that matters. The income is contractual rather than sentiment driven. The security is registered, first ranking and independently valued. And the loss mechanism requires two events jointly, a borrower default and a fall in the security's value through 35 points of equity, where an index fund requires only a change of mood. Our claim, stated exactly: not that the fund is insulated from the economy, but that it is exposed to a narrower and more underwritable slice of it, and that the order of losses is written into a mortgage rather than implied by a price.
What this paper does not claim: that a severe recession would leave the fund untouched. Arrears would rise, terms would extend, and in a deep enough downturn paired with defaults, losses are possible. The difference between us and an equity allocation in that scenario is the 35 cents of someone else's capital that burns first, and we would rather describe that mechanism precisely than borrow a word like uncorrelated to gesture at it.
Oxford Capital manages a fund that lends against Sydney property. Everything on this page is written by people whose livelihood improves if you find the asset class attractive. We cannot remove that conflict, so we manage it the only honest way available: primary sources for every figure, methods shown in full, and a standing commitment to publish the numbers that cut against us alongside the ones that flatter us. The Monitor above currently shows a falling market. It stays on the page.
Every number traces to the ABS, the RBA, APRA, Cotality, Bloomberg index data or an issuer's own documents, and the series or ISIN is printed beside it. Where we estimate, the working is in the appendix. Where sources disagree, for instance on the exact month of Sydney's recent peak, we say so rather than choosing the tidier answer.
We publish no price targets and no return projections, here or anywhere. We test loan structures against conditions that have already occurred, plus a margin, because history is harder to argue with than a model and considerably harder than a marketing department.
Corrections are appended to the original paper with a date, never silently edited in. Each paper also carries a short section titled "what this paper does not claim", which we regard as the most important paragraph in it.
The Sydney Mortgage Monitor and one paper, on the first business day of the month. No product marketing and no forecasts. The unsubscribe link is at the top of every email, where it belongs.
Research only. Nothing here is an offer of a financial product.