Census Tract Data for Real Estate Investors

A census tract is a small, relatively stable geographic subdivision of a county, designed by the U.S. Census Bureau to contain a population of roughly similar size and demographic character.

The practical implication is that tract-level analysis cuts through the noise that metro- or zip-code-level aggregation introduces. Two zip codes can share a boundary yet contain tracts with median household incomes that differ by tens of thousands of dollars, or with observed appreciation histories that diverge sharply over a decade. The Verus-AI model currently covers 82,179 scoreable U.S. census tracts spanning 3,234 counties, representing 84.09% of all U.S. tracts, which means the overwhelming majority of investable residential geography is captured in a single, consistently structured dataset. The 13 curated launch metros alone account for 7,233 scoreable tracts, a subset that illustrates how concentrated analytical depth can be even within a national framework.

One aspect that surprises many practitioners is how stable tract boundaries are between decennial censuses. The history window used in the model runs from 2014 to 2024, a span that captures both the post-financial-crisis recovery and the pandemic-era price acceleration, giving the historical appreciation component meaningful variation to work with.

The Verus-AI composite score runs from 0 to 100 and blends four components: the five-year forecast (40 points), 10-year historical appreciation (25 points), price stability (20 points), and income momentum (15 points). The weighting is intentional, forward-looking signal carries the most weight among the four components, but a tract that scores well on forecast alone while showing erratic historical behavior will be penalized by the stability component, which is derived from in-sample price residuals. A tract with a Low risk grade has tracked its model fit reliably; one rated High has not, regardless of where its composite score lands.

The forecast field, forecast_5yr_pct, projects the 2025-to-2029 cumulative change in median home value and is capped at 37.01%. That cap is not an artifact of conservatism; it is a deliberate model constraint that prevents the forecast from being read as a precise return target. When a tract hits the cap, the signal is directional, the model sees strong upward pressure, but the exact terminal value should be treated as an upper bound on the model's confidence, not a guaranteed outcome. The 80% confidence bands in the forecast chart widen materially over the five-year horizon, a feature worth examining: for the leading tract in the ranked table below, the terminal 80% interval spans $106,481, from $429,917 to $536,398, which is a wide enough range to matter for underwriting.

Income momentum, the income_yoy_pct field, is a smoothed year-over-year growth rate in median household income drawn from the American Community Survey 5-year estimates. Analysts should treat a strongly positive income_yoy_pct as a leading indicator worth monitoring rather than a mechanical price predictor.

The letter grade is a shorthand cut from the composite score, running from A+ down to F. It is useful for rapid screening across the ranked table below, but the underlying composite_score carries more information than the grade alone. Two tracts sharing a B+ grade can differ by nearly a full point in composite score, Tract 2509 in Phoenix carries a composite of 83.96 while Tract 1032 in Phoenix carries 83.42, both graded B+, which matters when ranking within a grade tier. Readers should consult the full numeric score before drawing conclusions from grade alone.

Tract 2219 in Phoenix (GEOID 04013422219) sits at the top of the ranked table among the 7,233 scoreable tracts across 13 major U.S. metros, carrying a Verus-AI score of 85, a grade of A-, and a Low risk rating. Its 2024 median home value is $350,500, and its median household income is $104,564, supporting a population of 3,690 residents. The income momentum reading of 9.83% is notably elevated relative to the distribution visible in the named_income_yoy_values array, where values of that magnitude appear but are far from universal, and where negative readings are common.

From a 2014 median value of $136,000, the tract reached $350,500 by 2024, a cumulative gain of 157.72% over the 2014-to-2024 history window. That is an uncapped, observed figure drawn from historical data, not a model projection, and it reflects the sustained price acceleration that characterized the Phoenix market across both the mid-cycle recovery and the pandemic-era surge. The forecast chart for this tract extends the series through 2029, with a median forecast path reaching $480,215 at the terminal year and an 80% confidence interval running from $429,917 to $536,398. The band width grows from $36,956 at the 2025 node to $106,481 by 2029, a pattern consistent with the compounding uncertainty inherent in any multi-year price model.

The forecast_5yr_pct for Tract 2219 is 37.01%, which is precisely at the model cap. That detail is analytically meaningful: the model's unconstrained signal was at or above the cap threshold, so the displayed figure represents a ceiling rather than a point estimate. Investors reading this tract's forecast should interpret it as the model expressing high directional conviction while acknowledging that the precise magnitude is bounded by the cap constraint. The Low risk grade reinforces the directional read, this tract has historically tracked its model fit well, but the widening confidence bands in the forecast chart are a reminder that even well-behaved tracts carry material uncertainty over a five-year horizon.

The Verus-AI data model draws from three primary federal sources and one proprietary layer. Tract boundaries and GEOIDs originate from the U.S. Census Bureau TIGER/Line and Gazetteer files, updated annually with decennial re-delineation. Population, income, rent, and education fields come from the Census Bureau's American Community Survey 5-year estimates, also updated annually. Observed median home value history for the 2014-to-2024 window is modeled from public price indices joined to the tract level, it is not a direct federal release but a derived series constructed to enable consistent longitudinal comparison. The five-year forecast for 2025 through 2029 and the Verus-AI composite score are proprietary model outputs refreshed on a per-model-refresh cadence.

The observed_appreciation_pct is a historical fact, 157.72% for Tract 2219 over 2014 to 2024, and carries the evidentiary weight of a measured outcome. The forecast_5yr_pct is a model-derived estimate subject to the assumptions embedded in the Verus-AI model, and the 80% confidence bands in the forecast chart are the appropriate lens for gauging its precision. Neither the Census Bureau nor any other federal agency produces the forecast; it is entirely a Verus-AI Research output.

The ACS income data deserves particular attention as a source limitation. Because the ACS 5-year estimates pool survey responses over a rolling multi-year window, the income figure attached to any given tract vintage reflects an average of conditions across several prior years rather than a single point-in-time snapshot. Analysts who weight income momentum heavily in their own frameworks should account for this lag when interpreting the field, particularly in markets where demographic change has been rapid.

The practical entry point for most investors is the ranked table, which surfaces the tracts with the strongest composite scores across the scored universe. Among the 7,233 scoreable tracts across 13 major U.S. metros, the top of the current ranking is concentrated in Phoenix and Dallas-Fort Worth, with Nashville and Austin also represented. Phoenix accounts for the majority of the top-12 positions, which is a pattern worth interrogating rather than accepting at face value: it may reflect genuine structural demand, or it may reflect the model's sensitivity to the strong observed appreciation that Phoenix posted over the 2014-to-2024 history window. The historical appreciation component carries 25 points in the composite, so markets with outsized recent history will tend to score well unless the stability and income components offset them.

The current_value field anchors the forecast to a real base. Tract 9205 in Dallas-Fort Worth carries a 2024 median value of $575,577, the highest among the 12 tracts in the ranked table, while Tract 9500 in Phoenix sits at $330,209, the lowest among that same set. Both carry forecast_5yr_pct readings at or near the 37.01% cap, but the absolute dollar exposure implied by those forecasts differs substantially. An investor underwriting at the $575,577 base faces a different capital requirement and exit liquidity profile than one underwriting at $330,209, even if the percentage signal is identical. The composite score does not adjust for absolute price level, so that dimension of risk must be layered in separately.

The named_income_yoy_values array illustrates the breadth of the distribution across scored tracts: readings range from deeply negative to strongly positive, with a substantial share clustered near zero or in modest negative territory. A tract showing a 9.83% income_yoy_pct, as Tract 2219 does, sits in the upper portion of that distribution and suggests demographic demand that has not yet fully capitalized into prices, which is the theoretical basis for the 15-point weight. However, the ACS smoothing effect means that a high reading today may partly reflect income conditions from two or three years prior, so the signal should be treated as confirming a trend rather than identifying a current inflection.

The risk grade provides a filter that the composite score alone cannot. A tract rated Elevated or High has shown wider residuals between its actual price path and the model's fitted path, meaning its historical behavior has been less predictable. For investors who require tighter underwriting assumptions, lenders, institutional buyers with mark-to-market exposure, the risk grade is a meaningful screen independent of the composite score. The forecast chart's confidence bands operationalize the same concept visually: a tract with a wide terminal band at the 80% level is telling the analyst that the model's own uncertainty about the outcome is high, and that the median forecast path should not be treated as a base case without acknowledging the full interval.