China real GDP growth nowcasting: Goldman Sachs’ China GDP tracker indicates 4.1% year-on-year growth in 2026 Q3, below its maintained 4.4% forecast.

Overview

Goldman Sachs presents a real-time framework for tracking China’s reported GDP growth amid weaker activity and more difficult post-pandemic forecasting conditions. Its current Q3 2026 nowcast is 4.1% year-on-year, below the institution’s maintained 4.4% forecast and implying downside risk pending late-quarter data.

Core views

China’s official real GDP growth slowed to 4.3% year-on-year in 2026 Q2 from 5.0% in Q1, a post-reopening low. First-half growth averaged 4.7%, broadly consistent with the official 4.5–5.0% full-year target range, but soft July and August activity prompted Goldman Sachs to cut its Q3 forecast to 4.4% year-on-year from 4.6%. The report sees the central questions as whether growth remains within the target range and whether policymakers add support if momentum weakens. It notes that the July Politburo meeting was more dovish than the April meeting. The institution argues that GDP forecasting has become materially harder since Covid. Bloomberg consensus’ mean absolute error for quarterly Chinese GDP growth has been roughly three times the pre-pandemic level. Goldman Sachs attributes this to unusually large year-on-year swings around the Covid shock and reopening, and to a changed policy reaction function: compared with the pre-Covid period, easing has become more measured, with a more muted and fleeting effect on growth. Historical relationships therefore provide less reliable guidance. Because China does not publish quarterly GDP by expenditure component in the detail available in major economies, Goldman Sachs builds a nowcasting framework from survey, trade and production data rather than replicating an expenditure-account approach. The model is estimated over 2017 Q1–2019 Q4 and 2023 Q2–2026 Q2, excluding the Covid and reopening period from 2020 Q1 to 2023 Q1. The report cautions that the effective sample is limited and out-of-sample performance remains an important caveat. The framework uses three models as the quarter’s information set expands. The earliest PMI-signal model regresses current-quarter GDP growth on official manufacturing and non-manufacturing PMIs. A bridge-indicator model, used after first-month data become available, combines non-manufacturing PMI, real export growth, auto-output growth and one-quarter-lagged GDP growth. The production-side model estimates GDP as a prior-year nominal-GDP-share-weighted average of agriculture, construction, industry and services; it uses industrial production and lagged industrial GDP for industry, and the services output index and lagged tertiary GDP for services. Agriculture and construction growth are carried forward from the prior quarter because timely data are limited; together they represent around 12% of nominal GDP. Goldman Sachs evaluates the models using RMSE, where a lower value denotes lower forecast error. Survey data offer an early signal, but accuracy improves through the quarter as hard data arrive, especially third-month releases given stronger quarter-end residual seasonality. The models outperform a simple AR(1) benchmark, and an equal-weighted average of all three beats individual specifications by reducing model-specific noise while incorporating both survey and hard-data signals. The live process begins with the PMI model, shifts to a two-model PMI/bridge average as trade and auto data arrive, and moves to an equal-weighted three-model average after first-month activity indicators are released around the middle of the second month. For 2026 Q3, the model currently points to 4.1% year-on-year real GDP growth, a further deceleration from Q2’s 4.3%. Goldman Sachs nevertheless maintains its 4.4% forecast because the latest model inputs do not fully capture possible incremental policy support, improved September weather and a recent tendency for stronger quarter-end activity even after seasonal adjustment. It cites faster government-bond issuance and reported implementation of an RMB800bn new policy-based financial instrument as signs that fiscal policy became incrementally more accommodative in September. Even so, the institution states that the nowcast implies downside risk to both its forecast and consensus expectations. It will reassess the tracker as the September 30 PMI release and October 14 trade data arrive.

Analysis framework

Goldman Sachs estimates three parsimonious GDP-tracking models using recent non-pandemic periods, then updates the applicable model mix as surveys, trade, auto production, industrial production and services data are released. It compares in-sample real-time accuracy using RMSE and adopts an equal-weighted model average after first-month activity data because that combination has the lowest error.

Methodology notes

• Other

  Real-time GDP nowcasting using PMI, bridge-indicator and production-side models

  The report updates an estimate of current-quarter reported GDP as monthly indicators arrive, rather than waiting for the official GDP release.

• Other

  Root mean squared error (RMSE) model evaluation

  Goldman Sachs compares forecast errors across model specifications; lower RMSE indicates a more accurate estimate and supports using the equal-weighted average.

Key data

• China real GDP growth, 2026 Q24.3% yoySlowed from 5.0% yoy in Q1.
• China real GDP growth, 2026 H1 average4.7%Described as broadly on track relative to the 4.5–5.0% full-year target range.
• Goldman Sachs 2026 Q3 GDP nowcast4.1% yoyBased on partial-quarter data; below Q2 growth and below the institution’s forecast.
• Goldman Sachs 2026 Q3 GDP forecast4.4% yoyLowered from 4.6% yoy after soft July and August activity data.
• New policy-based financial instrumentRMB800bnReported as implemented amid incrementally more accommodative fiscal policy in September.
• Agriculture and construction share of nominal GDParound 12%Their prior-quarter growth rates are carried forward because high-frequency tracking data are limited.

Impact & implications

The report views the current data as consistent with softer underlying Q3 momentum and a potential need for renewed policy support if weakness persists. Its maintained forecast depends partly on late-quarter factors not yet fully reflected in the model, while the model reading itself is a downside signal relative to both the Goldman Sachs forecast and consensus.

Risks

• The limited effective estimation sample means out-of-sample performance remains an important caveat.
• The historical model exercise is in-sample and backward-looking, so it is indicative of signal extraction rather than a definitive real-time track record.
• The current nowcast implies downside risk to Goldman Sachs’ 4.4% Q3 GDP forecast and to consensus expectations.

What to watch

• September PMI data due September 30.
• China trade data due October 14.
• Whether faster government-bond issuance and the reported RMB800bn policy-based financial instrument translate into stronger activity.
• Whether improved September weather and quarter-end residual seasonality lift growth beyond the current model signal.