GDP Nowcast: Q1 2026

Summary

Australia's GDP is tracking +0.78% quarter-on-quarter (+2.93% through the year) based on data available as at 6 April 2026. If realised, this would be the strongest quarterly outcome in three quarters and would push annual growth to its highest since early 2023.

This estimate combines seven active bridge models, each translating a monthly activity indicator into an implied GDP outcome, weighted by historical forecast accuracy. Labour market indicators are doing the heavy lifting; consumption is the principal drag.

Bridge estimates

The combined nowcast of +0.78% reflects the following individual bridge estimates and their inverse-MSE weights:

• Labour: hours worked – +1.03% Q/Q (weight: 31%)
• Labour: employment – +0.93% Q/Q (weight: 27%)
• Prices: monthly CPI – +0.88% Q/Q (weight: 8%)
• Investment – +0.84% Q/Q (weight: 8%)
• Trade – +0.65% Q/Q (weight: 10%)
• Production (Cobb-Douglas) – +0.10% Q/Q (weight: 12%)
• Consumption – −0.15% Q/Q (weight: 3%)

Labour: the dominant story

The two labour bridges together carry 58% of the combined weight. Hours worked is the single highest-weighted bridge at 31%, reflecting its historically tight relationship with output; employment follows at 27%. Both signal Q/Q growth above 0.9%, consistent with a labour market that has continued to add jobs and where average hours have ticked up in early 2026.

Consumption: the soft spot

The consumption bridge is the only model signalling contraction at −0.15%. Its weight is small (3%), so its drag on the combined estimate is limited. It nonetheless serves as a caution: households appear to have remained under pressure from elevated mortgage costs and cost-of-living fatigue. An upward surprise in consumption data would add materially to the nowcast given how far below the other bridges it sits.

Annual perspective

The through-the-year nowcast of +2.93% represents a full recovery from the 2024 trough of around 1.0% and is a material overshoot relative to the RBA's February 2026 SOMP forecasts, which had through-the-year growth peaking at around 2.5% before declining back toward 2% as the cash rate hikes took hold. At 2.93%, the nowcast is not just above that peak – it is above it at a point in the RBA's projected trajectory where growth was already supposed to be decelerating. 

With the NAIRU model estimating the economy's speed limit at around 2% per year, growth running nearly a full percentage point above that rate looks clearly inflationary, and is difficult to reconcile with the RBA's assumption that its tightening cycle would restore balance between demand and potential supply by mid-2028. The 70% confidence band around the Q/Q estimate runs from roughly +0.5% to +1.1%, derived from a backtest RMSE of 0.29 percentage points. A negative outcome is tail-risk territory given the strength of the labour signal.

Cautions

The current estimate should be read with some caution: as of 6 April the model is operating at T−2m, with only one month of within-quarter data available. This is the point in the information cycle where directional accuracy is weakest, as discussed in the backtest section below.

It is also worth noting that the one month of within-quarter data (from January 2026) predates the two RBA cash rate hikes delivered in Q1 2026. To the extent that tighter monetary policy dampens demand, the current estimate may overstate where the quarter lands – adding a downside risk that the headline number does not fully capture.

How the model works

The nowcast is built from eleven bridge equations – each one an OLS regression linking high-frequency indicators to quarterly GDP growth, with one lag of both the indicator and GDP growth included, and a COVID dummy covering 2020Q1–2021Q1. The eleven bridges fall into three groups: 

• six monthly bridges (that are always active), 
• four quarterly (which switch on progressively as the quarterly data are published), 
• and a production-function anchor that runs independently of the other bridges.

Monthly bridges (always active)

Six bridges use monthly ABS data, each SARIMA-completed to fill any missing months in the target quarter. As of 6 April, these are the only active bridges – the quarterly indicators have not yet been published for Q1 2026.

• Consumption – retail turnover (ABS 5682.0), aggregated as a quarterly sum, converted to log growth
• Investment – total dwelling approvals (ABS 8731.0), quarterly sum, log growth
• Labour: hours worked – monthly hours worked (ABS 6202.0), quarterly sum, log growth
• Labour: employment – employed persons (ABS 6202.0), quarterly mean, log growth
• Trade – balance on goods (ABS 5368.0), quarterly sum, used directly (not log growth, since the balance can be negative)
• Prices: monthly CPI – a spliced monthly CPI index combining the discontinued Monthly CPI Indicator (ABS 6484.0, from September 2017) with the current monthly series from ABS 6401.0 table 640106 (from April 2024 onwards), aggregated as a quarterly mean, log growth

The two labour bridges include an additional regressor: an HMA(13)-smoothed labour productivity trend derived from wage data (hourly compensation growth minus unit labour cost growth). This corrects for periods where employment grows faster than output – without it, the labour bridges systematically overestimate GDP during productivity downturns, introducing a positive bias of around 0.3 percentage points.

For SARIMA completion, a small candidate set of models is evaluated – (1,1,0), (1,1,1), (2,1,0), (0,1,1), each with and without a seasonal(12) component – and the best-fitting model by AIC is used to forecast the missing months. For the monthly CPI bridge, the genuine monthly observations are used for SARIMA rather than the quarterly-interpolated pre-2017 history, so the model learns real intra-quarter dynamics.

Quarterly bridges (not yet active for Q1 2026)

Four further bridges use quarterly indicators published in the weeks immediately before the GDP release. None of these is available yet for Q1 2026 – they will switch on progressively and revise the combined estimate as they arrive:

• Prices: CPI trimmed mean (ABS 6401.0 Appendix 1a) – published around five weeks before GDP
• Prices: WPI growth (ABS 6345.0) – published around three weeks before GDP
• Business: company profits growth (ABS 5676.0) – published around three days before GDP
• Business: business sales (ABS 5676.0) – published around three days before GDP

A bridge is excluded from the combination entirely if its target-quarter data are not yet published. The combination weights automatically renormalise across whatever bridges are active, so the estimate remains valid at any point in the information cycle.

Production bridge (Cobb-Douglas)

The final bridge is a Cobb-Douglas production function, independent of all the other bridges. It uses capital stock growth from the ABS Modellers Database (1364.0), labour force growth from the same source, and a time-varying capital share derived from factor income shares (gross operating surplus as a share of GOS plus compensation of employees). The multifactor productivity term is the Solow residual computed from wage data, smoothed with an HMA(51) filter but not floored at zero – negative MFP is real and informative for a nowcast of actual GDP, unlike in a potential output model where a floor would be appropriate. Its 12% weight reflects that it is somewhat less accurate than the labour bridges at this horizon, but it provides an anchor grounded in inputs fully independent of the other six monthly series.

Combination and uncertainty

Each active bridge is weighted by the inverse of its out-of-sample MSE, computed using an expanding window over the available history. The combined nowcast is the weighted average of the active bridge estimates. Prediction intervals are generated by bootstrap: residuals from all active bridges are pooled (scaled by their combination weights), resampled 1,000 times, and added to the combined point estimate. The 70% interval covers the 15th to 85th percentile of the bootstrap distribution; the 90% interval covers the 5th to 95th percentile.

Backtest performance (2022Q1–2025Q4)

The backtest uses the latest-revised data rather than true real-time vintages, which slightly flatters the results. With that caveat, performance across four information-set timings is as follows. At T−3m (previous quarter's GDP just published, no Q1 2026 data yet): RMSE 0.33%, bias −0.02%, direction accuracy 94%. At T−2m (one month of fast indicator data in): RMSE 0.38%, bias −0.20%, direction accuracy 81%. At T−1m (two months in): RMSE 0.33%, bias −0.09%, direction accuracy 88%. At T−0 (all monthly and quarterly data available): RMSE 0.29%, bias near zero, direction accuracy 94%.

The dip in directional accuracy at T−2m is worth noting. At that point the model has one month of fast indicators – employment and hours – but the slower indicators (building approvals, goods trade) are not yet in for the quarter, and SARIMA is completing two of the three months for most bridges. A single monthly observation can push the combined estimate the wrong way before the fuller picture arrives: the labour signal dominates with thin data, and if that first month is noisy, the direction call suffers. By T−1m a second month of labour data stabilises the estimate; by T−0 the quarterly bridges switch on and accuracy recovers to its T−3m level. The practical implication is that nowcasts based on a single month of within-quarter data should be read with more caution than those at either end of the information cycle.