Methodology, Build History & Full Disclosed Limitations

The index is built in two levels. Each of the six branches composes on its own components at equal weight; the index composes on the six branches at equal weight. Every reading below is expressed as a position from 0 to 100, and each is published on two frames.

The intensity frame reports how far a condition sits from its own historical centre, without regard to whether it favours investment. The signed frame applies a ratified direction to each component — a higher cost of capital counts against, a stronger approvals flow counts for — and reports the balance. A branch’s signed position is not the mirror of its intensity position, because a branch is signed through its components rather than as a whole: Supply Pipeline Conditions reads 98.4 intensity against 60.7 signed, not 1.6.

The six branch readings average to 0.6876 standard deviations above centre, which publishes as 90.4 on the intensity frame. Signed, they average to +0.0461 and publish as 53.4: the conditions lean slightly for investment. Both are indicative: the quarters in which all six branches are simultaneously at full component coverage number eleven, September 2023 to March 2026, below the twenty-quarter floor at which empirical ranking begins.

Two input sets are shared rather than separately acquired: the rate, yield and lending-standards inputs behind Financing & Volatility Conditions are mapped from Interest Rates (21210); the enforcement construction behind Policy & Regulatory Conditions is the one used by the APN Regulatory Velocity Multiplier™ (24210).

The signed composition as a quarterly series. The signed reading above takes each branch at its own latest quarter. The index also carries a quarterly signed series, in which each branch is signed and read in the same quarter, with the value of the dwelling stock excluded as in the signed reading. The APN Sovereign Policy Composite Index™ (24800) uses that series, inverted, as its investment-conditions term. At the March quarter 2026 it reads +0.0363σ, against the blended signed headline of +0.0461.

Restatement, 10 September 2026. The index previously published 87.4831 on a mean-of-ranks construction over six branches that composed directly on series. That reading is superseded by the construction described here.

An index built from six component branches can be assembled in two ways. It can rank each branch against its own history and then average the six ranks, or it can average the six readings and rank the result once at the end. The first discards magnitude. A branch sitting at the top of a narrow historical range and one sitting at the top of a wide range contribute the same number, and the distance between a branch at its centre and a branch at its outer range is reduced to the difference between two percentages.

The index takes the second route. Each branch reports a reading in standard deviations from its own historical centre; the index is the equal-weighted mean of those six readings, and the conversion to a published position happens once, at the end.

Three consequences follow, and each is why the construction was adopted.

Magnitude survives. A branch three standard deviations above its centre moves the index further than one a single standard deviation above it — the behaviour a reader expects, and the one a mean of ranks cannot produce.

The index acquires its own history. Because the composition is defined at every quarter rather than only at the current one, the index carries 242 quarters of composed history behind its published reading. Under the earlier certified construction, only the current reading was ever computed.

The conversion can be corrected for diversification. A mean of six readings varies less than any one of them unless the six move together perfectly. Converting that mean as though it carried the spread of a single reading would compress every index value toward the midpoint: on this index the uncorrected conversion reads 75.42 against a corrected 90.39, a difference of 14.97 points. The published position is therefore evaluated against the index’s own measured variability rather than an assumed one.

The ceiling behaviour the earlier construction was chosen to avoid is resolved rather than accepted. A rank against a record has no headroom once it reaches the highest value in that record, and reads 100.0 whether conditions sit at their previous peak or well past it. A reading in standard deviations carries no such limit, and the published position now separates the two — at index level. A component ranked empirically can still read at the top of its own record, and several do; the difference is that its reading in standard deviations continues to carry magnitude into the index, so a record high no longer flattens the headline.

The index now carries a composed value at every quarter in which at least one branch reports — 242 quarters in total. That history did not exist under the earlier construction, where only the current reading was ever computed.

It is not uniform. In any quarter where a branch is absent, the remaining branches are renormalised so their weights sum to one, and the composed value for that quarter rests on fewer than six. The quarters in which all six branches are simultaneously at full component coverage number eleven, September 2023 to March 2026. A chart of the full series would therefore place a three-branch quarter beside a six-branch one without distinguishing them, and the shape of the line would partly record changes in coverage rather than changes in conditions.

This is also why the published position is model-based rather than an empirical ranking: eleven quarters is below the twenty-quarter floor at which ranking begins. The floor is reached automatically as coverage lengthens, and the change of basis is disclosed in the quarter it occurs.

• Index

  Mean of component readings, converted once against the index’s own measured variability (Section 2 above).

• Index

  Position is indicative. Joint full coverage across all six branches stands at eleven quarters, September 2023 to March 2026, below the twenty-quarter ranking floor.

• Index

  Cross-quarter blending by design: 24710, 24740 and 24750 reference June 2026; 24720, 24730 and 24760 reference March 2026.

• 24710

  Six components correlated across two regulators, reading one macro-financial dimension from several angles. The high debt-to-income input carries 28 quarterly observations inside the baseline window, below the platform minimum, and sets this branch’s ranking window from the first quarter of 2019.

• 24720

  Both components hold their latest published value until the next publication, across a mixed six-monthly and annual cadence carried onto a quarterly grid, so consecutive quarters can read the same. Ranked against 45 quarters from March 2015; the March quarter 2026 reading equals the December quarter 2025 reading, the highest of the record, so the signed and intensity positions do not sum to 100.

• 24730

  The two components share an input by derivation — net overseas migration is itself a component of estimated resident population growth — and correlate at 0.9504. Equal weight retained; the overlap is disclosed rather than weighted away.

• 24730

  Both inputs enter as published by the ABS, without seasonal adjustment, and both run highest in the March quarter, when international student arrivals concentrate. The March-quarter reading has exceeded the preceding December-quarter reading in every year since 2012, so quarter-to-quarter movements in this branch are read against the same quarter of earlier years.

• 24740

  Four components drawn from the same two ABS collections, reading successive stages of one pipeline. The work-yet-to-be-done input is total building, residential and non-residential, deliberately wider than the residential-only measure used in the 21500 series.

• 24750

  The recorded source reference for the household spending input is inconsistent with the data and is under trace; the component’s direction does not depend on its outcome. The branch’s variance adjustment of 0.5527 sits below its independence bound of 0.5774. Flagged, not blocking.

• 24760

  Rents and Asset Base Value correlate above 0.97. Asset Base Value is excluded at zero weight from the signed index, and the signed position therefore carries no diversification adjustment. Rental inputs cover five of eight states and territories — Victoria excluded on methodology grounds, the Northern Territory and the Australian Capital Territory on data availability.

• 24763

  Rental Affordability is reserved, unmapped and inactive pending a denominator rebuild. It contributes nothing to any reading.