A Picture Is Worth 1,000 Words, Part II: Introducing A Unique Geospatial Signature for Economic Geography
Atlas Analytics’ Weekly Subscriber Update
TLDR:
Across 14 quarters of state-level analysis, Atlas ROY reveals a powerful insight: economic geography leaves a measurable spectral footprint.
States with clear physical activity, such as construction, energy, transportation, agriculture, show the strongest correlations with official GDP, while tech- and finance-heavy regions leave weaker or even inverted signals.
These differences form “Unique Geospectral Signatures for Economic Geography”, our term for how each region’s built and vegetative environment interacts with economic output.
The implication is profound: understanding these signatures may help explain not just why states diverge, but how entire economies develop.
What the Picture Says about Economic Geography
As we said before: A picture is worth a thousand words.
In part one of this three-part series, we explored how Atlas Analytics’ proprietary algorithm, ROY, works and what its capabilities are.
In Part II, we explore how Atlas Analytics’ GDP forecasts perform across the United States in comparison to the Bureau of Economic Analysis (BEA) official numbers, and what exactly that means for the U.S. economy.
In this article, we build towards something deeper and even profound: the discovery that every region leaves a unique geospectral signature, a distinct physical fingerprint that reveals how its economy grows, evolves, and behaves in real time.
This has important implications not only for the U.S., but for the economic development of every community and country around the world.
That’s Where Atlas Comes In
In last week’s article “The Road Ahead,” we discussed the ramifications of the government shutdown and how we still have no official visibility into Q3 economic activity, let alone Q4 (which is now more than half way over).
Government-produced GDP is slow, retrospective, and doesn’t tell us what is happening in real-time, and honestly, who has time for that in this economy.
That’s where Atlas Analytics comes in.
Using Atlas’ ROY, we can evaluate predictive strength across all states for the past 14 quarters, identifying where the model is most reliable, why it works best in certain regions, and what these patterns reveal about the structure of the U.S. economy itself.
This has an important implication: If we can understand where the model has worked well, we can gain insight about what it’s telling us for our predictions.
We will break down such predictions applied to all 50 U.S. states plus Washington D.C., comparing Atlas’ real-time satellite estimates to the official BEA GDP data, and what those patterns say about real-time economic activity.
A National Headshot
The heat map above represents the correlation between Atlas’ GDP predictions by state compared to the BEA’s GDP data between 2022 Quarter 1 and 2025 Quarter 2 (14 periods).
ROY is consistently most accurate in the Mountain West and the Central U.S., where correlations reach 0.42–0.46 regionally. ROY picks up real change here and matches GDP movements with impressive precision.
And what about the best performing states?
Interestingly, they are the more rural states ranging from Nevada in the west to Arkansas in the middle to New Hampshire in the north.
What the Patterns Reveal Across the Four Key Clusters
On a macro level, the U.S. is not just one economy: it is an amalgamation of regional agglomerations and ROY’s data is a mirror of that economic reality.
From our analysis, we view the predictions in the following four clusters:
Central States: AL, AR, IA, KS, KY, LA, MN, MS, MO, NE, ND, OK, SD, TN, TX
Coasts: AK, CA, CT, DE, DC, FL, GA, HI, ME, MD, MA, NH, NJ, NY, NC, OR, PA, RI, SC, VT, VA, WA, WV
Industrial Midwest: IL, IN, MI, OH, WI
Mountain West: AZ, CO, ID, MT, NV, NM, UT, WY
From a bird’s eye view, these maps below demonstrate both our prediction accuracy (RMSE) and precision (correlation)1:
With an average RMSE across all states of 3.5 and an average correlation of 0.3, Atlas predictions are often more accurate than the BEA’s own stated revision rate of 1.2 percentage points2.
We know ROY is exceptionally accurate where the economy is spatially visible, so we will continue to target regional versioning and variant testing to strengthen our forecasting in areas that are more complex.
Why Some Geographies Perform Well
Why does the Central U.S. perform best, while the neighboring Mountain West shows the lowest RMSE?
Why does Nevada exhibit our highest correlation, while California, the world’s fourth-largest economy, sits near the bottom?3
At a high level, the answer is simple: ROY performs best where the physical economy is tightly linked to economic output.
But the details matter:
The built environment is a weaker proxy for tech- or finance-driven GDP. Regions dominated by software, financial markets, biotech, and intangible services generate massive economic output with little visible physical change. Satellites excel at detecting infrastructure, logistics, and construction, but these sectors leave faint spatial footprints.
Housing dynamics vary dramatically across states.
In some regions, housing markets track employment, migration, and investment flows very closely; in others, the signal is noisy, reactive, or distorted by supply constraints. Where housing behaves like a clean leading indicator, ROY performs well. Where it does not, correlations weaken.Manufacturing-heavy states experience sharp, cyclical shocks.
Inventory swings, supply chain disruptions, auto production cycles, and plant downtime can produce sudden changes in output that may not register immediately in the built environment. ROY may pick up the direction correctly while missing the magnitude or timing.Economic complexity creates blind spots.
States with diversified or hybrid economies, mixing heavy industry with digital services, can produce conflicting signals, especially when one sector is booming while another contracts.
In short, regions with clear, continuous physical activity (energy, logistics, construction, transportation, agriculture, and industrial infrastructure) give ROY clean, high-fidelity signals. Regions dominated by intangible activity do not.
And that’s exactly why we do this work.
Atlas’ ROY model is a work in progress: improving, adapting, and learning every day as the U.S. economy evolves.
A Unique Geospectral Signature for Economic Geography
Each state interacts with the physical world in its own distinct way.
Construction patterns, industrial activity, and crop cycles all leave unique geospectral fingerprints in the satellite record, but states respond to these signals on different timelines.
Some geographies show long lags before physical change appears in economic statistics. Others exhibit surprising vegetation relationships, including occasional negative correlations, where increases in greenness coincide with economic slowdowns or vice-versa.
Why?
Our working hypothesis is that each geography blends its built environment, land use, and vegetative patterns differently, producing distinct spectral signatures that map onto local economic structures.
We call these Unique Geospectral Signatures for Economic Geography.
Understanding these signatures may help explain patterns of economic development writ large for global sustainable growth.
The Common Thread
What do the above patterns reveal about ROY’s GDP accuracy?
Rural states tend to have stable, structurally consistent economies — industries that evolve gradually and predictably over time.
They exhibit strong physical signals in the built environment: construction, infrastructure, transportation corridors, energy activity, and land-use patterns that satellites can track with precision.
They contain fewer “noise” variables from sectors like finance, software, or other intangible-output industries that leave little physical footprint.
And critically, they are not major financial trading hubs or “undetectable” technology hubs, where GDP growth is often decoupled from visible physical activity.
This common thread explains the broader regional patterns we see when we analyze the entire country through four distinct economic clusters. Taken together, these characteristics sharpen ROY’s visibility and allow the algorithm to excel where the real economy is spatially observable.
Where the real economy is spatially observable, ROY excels.
This common thread also explains the broader regional patterns we see when we analyze the entire country through four distinct economic clusters.
The Evolving Economic Landscape
That’s a wrap for Part II. In Part III, we will explore a time-series evolution of our state-level GDP forecasts versus the BEA, and what that portends for future economic development across the United States.
Curious to see the specs on your state?
Want our current state-level or national forecasts for Quarter 3 and Quarter 4? Contact us for a commercial, academic, and or institutional subscription.
This article was written with valuable research assistance from Morgan Reppert, Executive Operations Associate for Atlas Analytics.
What is RMSE? Root mean squared error is an accuracy measure that calculates the sign-indifferent average between our Atlas forecast and the actual BEA statistic.
This is the Mean Annual Revision (MAR) from first-release to the five-year annual revision for quarterly national-level GDP between 1999 and 2023 as stated by the BEA.
If California were a country, it would rank as the 4th-largest economy in the world behind only the U.S., China and Germany (in nominal terms, not PPP-adjusted).