AHI CPI Nowcast — —

Australia's only independent quantitative CPI nowcast. A multi-country ensemble model anchored to observed monthly ABS CPI data. Its pre-release accuracy, validation and methodology are published in full below. Most recent ABS release: —.

—
—
—
Annual CPI Inflation (ABS)
—
—
Primary sources only Multi-country ensemble Correct sign every quarter Pre-release accuracy published
Australia's Only Independent Quantitative CPI Nowcast

The AHI Nowcast publishes a real-time estimate of Australian quarterly CPI before the ABS official release — updated as each monthly observation arrives. Unlike bank forecasts, the full methodology, holdout validation, and error attribution are published transparently.

Pipeline  ·  Checking last pipeline run…
—  ·  resolved —  ·  STATUS: RESOLVED
Target quarter: —  ·  ABS release: —  ·  ■■■■■■■■ 100% resolved
Published before release
—
—
Actual (ABS index-derived)
—
—
Error
—
Direction
—
80% confidence band
—
ABS release date
—
Resolution status
—
Quarter coverage
100% resolved
Annual CPI (ABS)
—
Next prediction — pending data  ·  STATUS: AWAITING MONTHLY CPI ANCHOR (0 OF 2 RELEASES IN)

The platform's monthly-anchored methodology requires the first two ABS Monthly CPI releases within the target quarter before publishing a forecast.

The next quarter's nowcast will publish once the second monthly CPI release for the quarter has been ingested.

Next monthly CPI release
Pending — next ABS Monthly CPI release
ABS quarterly release
—
Pre-release accuracy — 2022-Q3 to 2025-Q4

Before each quarterly release we can observe at most two of the quarter's three monthly CPI figures, because the ABS publishes the third month at the same time as the quarterly CPI. These results show what the model would have published before each release, using only the monthly figures available at that point. For each quarter the model is re-estimated on earlier quarters only.

—
Mean absolute error
two monthly figures observed
—
Root mean square error
two monthly figures observed
—
Quarters inside the 80% band
two monthly figures observed
Monthly CPI observedQuartersMAERMSEBiasCorrect signAcceleration callInside 80% band
The forecast is published once two monthly figures are in. With one monthly figure the model's estimate is unchanged from the no-month estimate, because the anchoring step needs two.
Bias is the average of estimate minus ABS outcome. Correct sign means the estimate and the outcome were both positive or both negative. Acceleration call means the estimate correctly said whether quarterly inflation would rise or fall from the previous quarter. Model inputs are the data as held today, not the earlier versions available at the time.
Published forecasts

Each row shows the forecast published on this page before the ABS release, and the ABS outcome.

QuarterPublished before releaseBasisActual (ABS)ErrorAcceleration callPost-release rerun
Error = estimate minus ABS outcome (positive = over-predicted). Actuals are derived from the ABS quarterly CPI index. A post-release rerun is the current model run after the release with all three monthly figures. It is shown for reference only: it is not a forecast and is not used in the error column.
Monthly-to-quarterly reconciliation — 14 quarters, 2022-Q3 to 2025-Q4

This table runs the same model with all three monthly figures in hand. Those figures arrive at the same time as the quarterly CPI, so this measures how closely the monthly series adds up to the quarterly result. It is not a test of forecasting. Mean absolute error —, root mean square error —. Correct sign in — quarters; acceleration called correctly in —.

QuarterAHI EstimateActual (ABS)ErrorAcceleration call
AHI Estimate = model estimate including policy calendar adjustments. Error = AHI minus ABS (positive = over-predicted), computed from full-precision values and rounded independently of the estimate and actual columns.
† Headline CPI suppressed by government electricity rebates (−71bp, −57bp). Ex-policy underlying CPI: +0.93%, +0.77%.   ‡ The ABS Monthly CPI Indicator diverged from quarterly headline CPI by 0.38pp in Q1 2024 because the monthly basket differed from the full quarterly CPI. 2022-Q3 uses the quarterly index because July 2022 monthly CPI is not held.
Confidence band

The 80% band shown with each forecast is a fixed width around the estimate.

Supply Shock Quarters — 2022-Q1 and Q2

In the first two quarters of 2022 the ABS had not yet launched its monthly CPI, so no monthly figures were available and the model ran on its own. These quarters are reported separately from the results above. The platform now publishes a forecast only once two monthly figures for the quarter are in.

The 2022 supply chain disruption had no precedent in the model's training data: the NY Fed Global Supply Chain Pressure Index reached 4.3 standard deviations above its long-run mean. Monthly anchoring, available from 2022-Q3 onward, addresses this.

Model signal categories
Official Government Data
ABS price indices, RBA monetary data, ABS Wage Price Index, ABS Import Price Index, ABS Producer Price Index, and ABS Monthly CPI Indicator — used for monthly CPI anchoring as confirmed releases become available.
Financial Market Signals
International crude oil benchmarks and currency exchange rates providing import price and commodity shock signals. BIS credit-to-GDP gap measuring household and corporate leverage across six comparable economies.
Supply Chain Signals
NY Fed Global Supply Chain Pressure Index monthly from 1997. IMF Primary Commodity Price Index quarterly from 1960. US ocean freight price index from 1986. Regime classifier activates supply-shock parameters when GSCPI exceeds 1.5 standard deviations above its long-run mean.
AHI Proprietary Signals
AHI Stress Cycle Gap — proprietary measure of the distance between cost pressure and composite household stress. Validated as an inflationary persistence indicator in inflationary macro regimes. Balance sheet vulnerability from RBA debt service ratios and ABS household saving data. Unit labour costs from ABS national accounts.
Methodology

The AHI Nowcast is an ensemble forecasting model combining multiple regression and time-series techniques, with a gradient-boosted corrective element. Features are selected from monetary policy signals, price momentum indicators, import cost transmission channels, and the AHI stress cycle gap — a proprietary measure of household cost absorption with demonstrated inflationary persistence properties. Calibrated against two decades of international comparable-economy data. Specific model architecture, feature selection, ensemble composition, and training parameters are proprietary.

Monthly CPI anchoring incorporates confirmed ABS Monthly CPI Indicator data as quarters progress. Fuel prices use daily retail pump price data (population-weighted NSW×60%, QLD×40%), replacing wholesale TGP — consistent with the ABS CPI motor fuel methodology. Policy adjustments are applied from the AHI policy calendar (tobacco excise, electricity rebates, PBS changes). The all-in AHI estimate is compared against the ABS official CPI release.

Walk-forward validation protocol

For each evaluation quarter the model is re-estimated using only earlier quarters, then run with the monthly CPI figures the ABS had published before that quarter's release: none, the first month, or the first two months. The third month is published together with the quarterly CPI, so it is never available beforehand. The model's inputs are the data as held today rather than the earlier versions available at the time, so these results are a close approximation to real-time accuracy, not an exact replay.

The monthly-to-quarterly reconciliation runs the same model with all three months. It shows how closely the monthly series adds up to the quarterly result, not how well the model forecasts. The 2022 supply shock quarters, before the ABS monthly CPI existed, are reported separately. The evaluation is regenerated from the platform's code and data every week and checked automatically against the published figures.

Disclaimer. This nowcast is produced independently by the Australian Household Index and is not affiliated with, endorsed by, or connected to the Australian Bureau of Statistics (ABS), the Reserve Bank of Australia, or any government agency. It is provided for informational and research purposes only and does not constitute financial, investment, or economic advice. Past predictive accuracy is not a guarantee of future performance. The AHI CPI Nowcast will diverge from the official ABS CPI release — sometimes materially — particularly during structural breaks, policy changes, or data revisions.