The four signals behind Australia’s cost of capital
The cash rate gets the headlines, but the cost of capital in Australia is not one number — it is a system of at least four, and reading only the headline rate is how most commentary goes wrong. Interest Rates (21210) is the APN Codex node built to read all four together, each against its own fifteen-year history, so that “rates are high” stops being an impression and becomes a measurement.
The four vectors
1. The cash rate itself — the RBA’s policy setting, resampled to quarterly observations for baseline purposes (60 observations in the certified baseline). Its fifteen-year average is 2.36%; the certified June 2026 reading of 4.35% sits well above it. This is the wholesale price of money.
2. The retail lending rate — what banks actually charge home borrowers, from the RBA’s F5 indicator series (60 quarterly observations, average 6.24%). Certified June 2026: 8.77%. This is the price households really pay, and the roughly 4.4-percentage-point spread over the cash rate is one of the most stable facts in the whole record.
3. The 2-year government bond yield — a weekly series, aggregated to 50 quarterly observations for baseline purposes, capturing where financial markets expect rates to travel over the next couple of years. Its baseline average is 2.13%; the certified June 2026 reading stands at 4.41%. When this vector runs above the cash rate, markets are betting conditions stay restrictive.
4. High debt-to-income lending — the share of new mortgages written to borrowers taking on debts of six times income or more, from APRA’s quarterly statistics (28 observations). Baseline average 12.5%; the certified June 2026 reading was 5.40%. This vector measures how much risk the banking system is actually writing.
How the Codex turns four series into one reading
Each vector is converted into a Z-score: its distance from its own fifteen-year average, measured in units of that series’ normal variation (its standard deviation, written σ). A Z-score of 0 means “exactly at the long-run average”. +1σ means one standard deviation above — noticeably elevated. Beyond ±2σ is statistically abnormal territory. The advantage of the Z-score is comparability: a percentage rate, a bond yield and a lending share can all be read on the same axis. The node’s composite is simply the equal-weighted average of the four vector Z-scores, computed once all four vectors report for the same quarter.
The prior certified composite, for March 2026, was +1.00σ. A quarter is certified only once all four vectors have reported for it; the September 2026 quarter waits on APRA’s September edition, listed for December 2026. Coverage is disclosed explicitly rather than assumed.
What the configuration is saying
All four vectors sit at least one standard deviation from their fifteen-year norms, and three of the four sit above +1σ simultaneously — with the highest of the three being the bond market’s forward view. That ordering matters: it means the restrictive setting is not just the current policy stance but the market’s expectation for the medium term. Meanwhile the fourth vector, high-DTI lending, sits below its historical average (−1.07σ), consistent with APRA’s 20% cap on high-leverage lending, in effect since February 2026, continuing to bind. Put together: money is expensive, expected to stay expensive, and the riskiest form of borrowing has been squeezed to historically low levels by regulation on top of price.
Terms worth keeping
Baseline — the frozen fifteen-year window (2011–2025) each series is measured against; it does not drift as new data arrives. Certified — a reading formally locked into the APN database as the historical record, as distinct from live readings that may still be revised. NIL RETURN — the Codex’s rule that a missing number is disclosed as missing, never quietly filled in. Coverage — how many of the composite’s vectors were actually present for a given period; always check it before comparing two periods.
The forward yield vector, isolated
The four-vector list above introduced the 2-year government bond yield as the vector that “captures where financial markets expect rates to travel.” It’s worth isolating why this particular vector behaves differently from the other three, and why it matters more than its 25% weighting in the composite might suggest.
The cash rate and the retail lending rate are both administered prices — one set by the RBA, the other set by banks pricing off the RBA’s setting plus their structural margin. The forward yield is a market price: it’s set continuously by investors buying and selling government bonds, and it reflects their collective bet on where policy is headed, not where it currently sits. This is why the forward yield vector can diverge from the cash rate vector in a way the retail rate vector cannot — the retail rate is contractually tethered to the cash rate through the bank’s margin, but the bond market is free to price in expectations the RBA hasn’t yet acted on.
The certified record illustrates this directly. In March 2026 the forward yield vector’s +2.17σ sat well above the cash rate vector’s +1.20σ, a gap of 0.98 standard deviations and the widest since September 2022. By June the gap had narrowed to 0.48 (+1.85σ against +1.37σ), because the RBA raised the cash rate in May while the bond yield fell. The gap measures how much further tightening the market is pricing beyond the current setting. A narrowing can come from either side: from the yield falling, or from the cash rate catching up to it. In the June quarter it was both, and the RBA raised the cash rate again on 30 September.
