The only free source that answers "who actually lives here" at the tract level. A Census skill teaches Claude which ACS variables underwrite renter-market calls, income-to-rent burdens, and tenure shifts — the reads that separate a real submarket thesis from a metro-level guess.

Reference US demographics, household income, and housing-stock data from Census (U.S. Census Bureau, American Community Survey 5-Year Estimates) for real estate investment decisions. Use this skill whenever a user asks about ACS, ACS 5-Year, Census tract data, block-group demographics, median household income, median home value, median gross rent, median owner costs, rent burden, homeownership rate, vacancy rate, housing units, year built, population growth, educational attainment, bachelor's-degree share, labor force participation, B25 housing tables, B19013, B25064, B25077, county-grain demand benchmarks, or which Census variables underwrite a residential RE decision. Maps 12 ACS 5-Year metrics across 5 categories (demographics, income_affordability, housing_costs, tenure_vacancy, housing_stock) to investor use-cases. Progressive disclosure — decision playbooks and the full 12-dataset reference live in `references/`.

Census for Real Estate

What is Census

Census is the U.S. Census Bureau's public data portal — data.census.gov. Free, API-keyed, and the only free source that publishes household-level demographics, income, and housing-stock data down to block-group grain. For US residential real estate, it's the authoritative single stop for tract-level demand reads — population, income, education, tenure, rent burden, and housing stock — that no listing feed or macro source covers.

This skill targets the American Community Survey (ACS) 5-Year Estimates specifically. Each vintage is a 5-year pooled estimate released annually each December; the 5-Year (not 1-Year) flavor is the only one that fills at every geography investors cite — counties under 65k population and every tract simply don't appear in 1-Year. For current-cycle pricing and listings, pair Census with Redfin, Realtor, or Zillow — Census is slower than the listing feeds but goes finer than any of them.

Bundled resources

Progressive disclosure rule: start with this file and the category table below. Read references/metrics.md only when a specific dataset or dataset_key is actually needed. Read the YAML only when the markdown is insufficient.

Getting an API key

1. Request a free key at api.census.gov/data/key_signup.html. Arrives by email in minutes.
2. Set environment variable: CENSUS_API_KEY=your_key_here
3. No dedicated client needed — requests and pandas cover every pull. A pip install census wrapper exists if you want one.
4. Rate limit: 500 requests/day per IP without a key, effectively unlimited with one. Get the key.

Minimal Python example

import os
import requests
 
CENSUS_API_KEY = os.environ["CENSUS_API_KEY"]
 
# Median household income — Hillsborough County, FL — ACS 5-Year 2024
url = (
    "https://api.census.gov/data/2024/acs/acs5"
    "?get=NAME,B19013_001E&for=county:057&in=state:12"
    f"&key={CENSUS_API_KEY}"
)
header, row = requests.get(url, timeout=30).json()
print(row[0], int(row[1]))

Swap B19013_001E for any ACS variable from references/metrics.md.

The 5 categories

Total: 12 datasets. See references/metrics.md for the full list with dataset_keys and ACS variable IDs.

Decision playbooks

Concrete "the user asked X — here's the dataset-key set to pull." Use these as starting points; extend when the question demands it. If a playbook names a dataset, references/metrics.md has the verbatim rationale.

"Can renters here afford a $2,000/mo unit?" → census_median_income, census_median_gross_rent, census_median_rent_burden Apply the 30%-of-income rule to median income for the local rent ceiling, then read gross rent and the rent-burden median to see how much room is left.

"Is this market rent-vs-buy biased?" → census_median_gross_rent, census_median_owner_costs, census_median_home_value, census_homeownership_rate Symmetric pair: rent vs all-in mortgaged carry. Add home value and the owner share to read whether the market structurally favors one tenure.

"What does the demand side look like?" → census_population, census_bachelors_or_higher, census_labor_force_participation Population growth is the demand denominator; bachelor's-share momentum signals high-income inflow; LFPR says how many adults can actually pay rent.

"Is this market tight or loose?" → census_vacancy_rate, census_homeownership_rate, census_housing_units Vacancy under ~7% = tight (rents push, leasing compresses); above ~12% = loose. Tenure mix and stock scale add context.

"Is the housing stock aging or expanding?" → census_housing_units, census_median_year_built YoY housing-units growth is net new construction. Median year built drifting older = construction lagging growth, value-add opportunity widens.

"Is this market priced to local income or imported demand?" → census_median_income, census_median_home_value, census_median_gross_rent Benchmark the county to FL + US on each. Income-aligned pricing = local-market; price and rent decoupled from income = exogenous demand (in-migration, vacation, institutional capital).

Gotchas

• 5-Year ≠ a single year. ACS 5-Year vintages are pooled 5-year averages. The 2024 vintage covers 2020–2024, not just 2024. Don't compare a 5-Year reading to a 1-Year or single-year point estimate without normalizing the window.
• Annual release lag. New ACS 5-Year vintages drop every December. Through Q3, the most recent vintage is ~9–21 months stale relative to the current cycle.
• 1-Year doesn't exist for most counties. Anything under 65k population and every census tract only ships in the 5-Year flavor. For a freshest single-year estimate at sub-metro grain, you can't get it from Census — go to listing-feed sources.
• Self-reported home values. B25077 is the owner's estimate of market value, not a transaction price. Lags the real cycle by the 5-year pooling window; cross-check Realtor / Zillow / Redfin for current-cycle marks.
• Inflation-adjusted to vintage end-year. B19013 dollars are real-to-the-vintage. Year-over-year comparisons across vintages are nominal — adjust separately for real growth.
• Margin of error widens at smaller geographies. County is the safe default. Tract and block-group MOEs can run ~20% of the estimate; don't push a single-tract reading without checking the published MOE.
• Gross rent (B25064) ≠ contract rent (B25031). Gross includes utilities, contract doesn't. Most investor comps quote gross — match what your comps use.
• Derived metrics need the derivation. Homeownership rate, vacancy rate, bachelor's-share, and LFPR are computed from raw variables. The catalog's derivation block specifies the formula — don't pull a single variable and expect a percentage.

Source of truth

This skill is derived from TiRE's canonical catalog at agents/data/census-datasets.yaml. The YAML is hand-curated; every entry has a why_it_matters rationale that feeds both this skill and TiRE's research agents.

When the upstream catalog changes, regenerate references/metrics.md and references/census-datasets.yaml. v1 is hand-curated; build script deferred until drift is observed.