When Headline GDP and Core Economic Activity Diverge
A decomposition of the Q4 2025 GDP release — and what it reveals about trade, inventories, and real-time measurement
This week, the BEA released its first estimate of Q4 2025 GDP at 1.4%, below our Atlas Analytics forecast of 5.2%.
We believe in addressing misses directly.
Forecasting quarterly GDP, particularly at the subcomponent level, is one of the most complex measurement challenges in macroeconomics. This release provides an opportunity to examine where divergence occurred and what it reveals about the structure of GDP itself.
Below is a decomposition of the variance.
Breaking Down the Gap
GDP is the sum of three major contribution buckets:
Core GDP (Consumption + Government + Fixed Investment)
Private Inventories
Net Exports
Our variance relative to the BEA print was concentrated in two areas: Net Exports and Inventories, with an additional meaningful drag from Government spending that downwardly affected Core GDP.
1. Core GDP
Atlas forecasted 2.35pp versus the BEA’s 1.13pp.
The primary divergence within Core GDP came from Government spending, which subtracted approximately 0.9pp from growth following last fall’s government shutdown. That contraction meaningfully reduced headline growth and amplified the overall gap.
Importantly, private-sector activity signals (consumption + fixed investment) remained materially stronger than the headline figure suggests, and if Government spending had remained consistent with previous quarters instead of the protracted shutdown, the gap between Atlas’ Core GDP and the actual would have been negligible.
2. Private Inventories
Atlas projected 0.77pp versus the BEA’s 0.21pp.
While directionally consistent, inventories contributed less to growth than anticipated. Inventory accounting remains one of the most volatile components of quarterly GDP and is often revised in subsequent releases.
3. Net Exports
Atlas forecasted 2.05pp versus the BEA’s 0.08pp.
This was the largest source of variance.
Trade flows, particularly imports, can swing quarterly GDP materially due to timing effects, booking conventions, and seasonal adjustments. Measured trade activity and its contribution to GDP are not always perfectly synchronized in real time.
This quarter highlighted that volatility clearly.
Measurement Is Hard — Especially at the Margin
Quarterly GDP is constructed from thousands of data inputs, surveys, and accounting adjustments. Certain components, particularly Net Exports and Private Inventories, are inherently more volatile and prone to revision.
This release reinforces three realities:
Subcomponent forecasting is significantly more difficult than headline alone.
Trade and inventory dynamics can overwhelm otherwise stable private-sector growth.
Real-time measurement requires continuous refinement.
What We Are Doing
In response, we are:
Operationalizing additional algorithms and model modules focused specifically on volatile trade and inventory dynamics.
Continuing to refine our Core GDP tracking, where underlying private-sector signals remain consistent, while expanding our capacity to incorporate atypical government spending shocks and other non-recurring economic events.
Conducting a full post-release decomposition and lag analysis.
Our objective is not short-term perfection.
It is a durable measurement improvement.
The Broader Signal
Despite the headline print, our Core GDP signal continues to indicate steady underlying economic activity.
Quarterly prints can be noisy.
Economic structure is not.
Understanding the difference between the two is precisely why high-frequency, observational data matters.
Transparency as Our Policy
We will continue to publish both our successes and our misses.
Forecasting is probabilistic. Measurement is iterative. Credibility comes from transparency.
More detailed analysis, including component-level diagnostics and model adjustments, will follow and we will be releasing updates to our key algorithms (including new ones) in the weeks to come.
As always, we appreciate your engagement and thoughtful dialogue.
Is there a way to build political factors (party in power, size of majority across houses, year in election cycle, and likewise for major trading partners) into the model, as presumably these will have a significant impact on the Core GDP?