⚙️ How-To Building permits & new housing Data sources & downloads Notebook By

How to Read Census Building Permits Survey Data in Python

On this page

Problem statement

The US Census Bureau counts the new homes that local governments authorise every month: each permit-issuing place reports how many buildings they permitted, how many housing units those buildings hold and what they are expected to cost to build. The county files are the easiest way into it, one row per county, every year back to 1990. They are plain comma-separated text, but with two header rows, a blank line, 30-character padded names and two parallel sets of counts, so a plain read_csv gets almost everything wrong. This page reads them properly, with the 2025 file and all 36 annual files, and shows what the rows contain.

Quick answer

Skip the two header rows and the blank line, name the 30 columns yourself, and read the codes as text. The setup below wraps that in permits():

y25 = permits("co2025a.txt")
print(y25[["GEOID", "name", "units", "value"]].sort_values("units", ascending=False).head(5).to_string(index=False))
GEOID               name  units      value
04013    Maricopa County  31651 8516645698
48201      Harris County  30909 7651942703
06037 Los Angeles County  22250 4573893701
48085      Collin County  19082 4745418498
12086  Miami-Dade County  16535 4730017820

Maricopa County, Arizona, permitted the most housing units in 2025, 31,651, just ahead of Harris County, Texas, with 30,909. value is the declared construction cost in dollars: $8.5 billion in Maricopa.

Horizontal bar chart of the twelve counties that permitted the most housing units in 2025, led by Maricopa County, Arizona with 31,651 and Harris County, Texas with 30,909, then Los Angeles, Collin, Miami-Dade and Travis counties. Horizontal bar chart of the twelve counties that permitted the most housing units in 2025, led by Maricopa County, Arizona with 31,651 and Harris County, Texas with 30,909, then Los Angeles, Collin, Miami-Dade and Travis counties.
The counties that permitted the most housing units in 2025.

Header rows, sizes and counties

1. Look at the first lines

for line in Z.open("co2025a.txt").read().decode().splitlines()[:4]:
    print(line[:118])
Survey,FIPS,FIPS,Region,Division,County,,1-unit,,,2-units,,,3-4 units,,,5+ units,,,1-unit rep,,,2-units rep,,,3-4 unit
Date,State,County,Code,Code,Name,Bldgs,Units,Value,Bldgs,Units,Value,Bldgs,Units,Value,Bldgs,Units,Value,Bldgs,Units,V
 
2025,01,001,3,6,Autauga County                ,200,200,82131895,0,0,0,0,0,0,0,0,0,200,200,82131895,0,0,0,0,0,0,0,0,0

The first line names the size of building each group of three columns describes, the second line names the three columns (buildings, units, value), and the third is blank. The rep groups repeat the counts for reported permits only; the first four groups are the Bureau's estimates, which fill in places that did not report. Every county name is padded with spaces to 30 characters.

2. Units by size of building

nat = pd.DataFrame({"buildings": [y25[f"est{s}_bldgs"].sum() for s in SIZES],
                    "units": [y25[f"est{s}_units"].sum() for s in SIZES]},
                   index=["1 unit", "2 units", "3-4 units", "5+ units"])
nat["share of units"] = (nat.units / nat.units.sum()).map("{:.1%}".format)
print(nat.to_string())
           buildings   units share of units
1 unit        911903  911903          63.7%
2 units        18375   36748           2.6%
3-4 units       5268   18993           1.3%
5+ units       16414  463972          32.4%

Of the 1,431,616 units permitted in 2025, 63.7% were single-family houses and 32.4% were in buildings of five or more units. The five-plus group is where buildings and units part ways: 16,414 buildings held 463,972 units.

Two bar charts by size of building for 2025: buildings are almost all single-family, 911,903 against 16,414 buildings of five or more units, while units split 911,903 single-family against 463,972 in five-plus buildings. Two bar charts by size of building for 2025: buildings are almost all single-family, 911,903 against 16,414 buildings of five or more units, while units split 911,903 single-family against 463,972 in five-plus buildings.
Few apartment buildings, many apartments: 16,414 five-plus buildings held 463,972 units.

3. Rows that are not counties

odd = y25[~y25.GEOID.isin(pop.GEOID)]
print(odd[["GEOID", "name", "units"]].to_string(index=False))
missing = pop[~pop.GEOID.isin(y25.GEOID)]
print(f"counties with no row: {len(missing)}, population {missing.POPESTIMATE2025.sum():,}; rows with zero units: {(y25.units == 0).sum()}")
GEOID                        name  units
23000 Maine Unorganized Territory    114
26000   Michigan Balance of State    135
30000    Montana Balance of State      0
35000 New Mexico Balance of State   2752
41000     Oregon Balance of State    169
counties with no row: 120, population 797,240; rows with zero units: 248

Five rows have county code 000. They are permits from places the Census Bureau does not assign to a single county: Maine's unorganized territory and "Balance of State" rows for Michigan, Montana, New Mexico and Oregon. New Mexico's holds 2,752 units, enough to matter in a state total but impossible to map. In the other direction, 120 counties, home to 797,240 people, have no row at all, because no place in them issues permits, and 248 rows are present with zero units. A missing county is unknown, not zero.

States, years and a county map

The data used on this page

The US Census Bureau's Building Permits Survey county files (annual 1990–2025 and year-to-date 2025–2026), its Northeast place files for 2020–2025, the Vintage 2025 county population estimates and the 2024 cartographic boundary counties, all public domain. They were downloaded on 16 October 2026 and saved on this site in one archive, because Census revises recent months as late reports arrive.

from pathlib import Path
import zipfile

import geopandas as gpd
import numpy as np
import pandas as pd
import requests

Path("data").mkdir(exist_ok=True)
# US Census Bureau Building Permits Survey: annual county files co1990a.txt-co2025a.txt, year-to-date county files
# co2501y.txt-co2512y.txt and co2601y.txt-co2608y.txt, Northeast place files ne2020a.txt-ne2025a.txt; the Vintage 2025
# county population estimates; 2024 cartographic boundary counties. Downloaded 16 October 2026, saved here unchanged.
# Live: https://www2.census.gov/econ/bps/County/co2025a.txt (and the same folder for the others),
# https://www2.census.gov/econ/bps/Place/Northeast%20Region/ne2025a.txt,
# https://www2.census.gov/programs-surveys/popest/datasets/2020-2025/counties/totals/co-est2025-alldata.csv,
# https://www2.census.gov/geo/tiger/GENZ2024/shp/cb_2024_us_county_20m.zip
ARCHIVE = "census-building-permits-2026-10-16.zip"
if not Path(f"data/{ARCHIVE}").exists():
    Path(f"data/{ARCHIVE}").write_bytes(requests.get(f"https://data.spatialworkflow.io/{ARCHIVE}", timeout=600).content)
Z = zipfile.ZipFile(f"data/{ARCHIVE}")

SIZES = ["1", "2", "34", "5"]  # buildings with 1 unit, 2 units, 3-4 units, 5 or more units
COLS = (["survey", "state", "county", "region", "division", "name"]
        + [f"{k}{s}_{m}" for k in ("est", "rep") for s in SIZES for m in ("bldgs", "units", "value")])


def permits(name):
    """One county file: two header rows and a blank line, then a row per county. est = estimate with
    imputation for places that did not report, rep = reported only."""
    d = pd.read_csv(Z.open(name), skiprows=3, header=None, names=COLS, skipinitialspace=True,
                    dtype={"survey": str, "state": str, "county": str, "name": str})
    d["survey"], d["name"] = d.survey.str.strip(), d.name.str.strip()
    d["state"], d["county"] = d.state.str.strip().str.zfill(2), d.county.str.strip().str.zfill(3)
    return d.assign(GEOID=d.state + d.county,
                    units=d[[f"est{s}_units" for s in SIZES]].sum(axis=1),
                    reported=d[[f"rep{s}_units" for s in SIZES]].sum(axis=1),
                    value=d[[f"est{s}_value" for s in SIZES]].sum(axis=1))


pop = pd.read_csv(Z.open("co-est2025-alldata.csv"), encoding="latin-1", dtype={"STATE": str, "COUNTY": str})
pop = pop[pop.SUMLEV == 50].assign(GEOID=lambda p: p.STATE + p.COUNTY)
Path("data/cb_2024_us_county_20m.zip").write_bytes(Z.read("cb_2024_us_county_20m.zip"))
counties = gpd.read_file("data/cb_2024_us_county_20m.zip")
counties = counties[~counties.STUSPS.isin(["AK", "HI", "PR"])].to_crs(5070)
y25 = permits("co2025a.txt")
print(f"2025 county file: {len(y25):,} rows, {y25.units.sum():,} units permitted; {len(pop):,} counties in the estimates")
2025 county file: 3,029 rows, 1,431,616 units permitted; 3,144 counties in the estimates

Example 1: units by state

st = y25.merge(pop[["GEOID", "STNAME"]], on="GEOID").groupby("STNAME").units.sum().sort_values(ascending=False)
print(st.head(6).to_string())
print(f"top 3 states: {st.head(3).sum() / y25.units.sum():.1%} of all units")
STNAME
Texas             210217
Florida           178297
California        103856
North Carolina     86167
Georgia            61196
Arizona            51532
top 3 states: 34.4% of all units

Texas, Florida and California permitted 34.4% of all units in 2025; Texas alone 210,217. The state comes from the first two digits of GEOID, so the names here come from the population estimates, which spell every state out.

Example 2: all 36 annual files in one frame

years = pd.concat([permits(f"co{y}a.txt").assign(year=y) for y in range(1990, 2026)], ignore_index=True)
by_year = years.groupby("year").agg(rows=("GEOID", "size"), units=("units", "sum"))
print(by_year.loc[[1990, 2005, 2009, 2025]].to_string())
print(f"survey codes: {years.groupby('year').survey.first().loc[[1990, 1998, 1999]].tolist()}")
      rows    units
year               
1990  2994  1118551
2005  3027  2155316
2009  3026   582963
2025  3029  1431616
survey codes: ['9099', '9899', '1999']

The 36 files stack into one frame with a year column. The national total runs from 2,155,316 units in 2005, the peak of the housing boom, to 582,963 in 2009. Before 1999 the survey date is written as 9099, 9199 and so on through 9899, so take the year from the file name, not from that column.

Example 3: a county map

m = counties.merge(y25[["GEOID", "units"]], on="GEOID", how="left")
print(f"counties drawn: {len(m):,}; with a permit row: {m.units.notna().sum():,}; 1,000+ units: {(m.units >= 1000).sum()}")
ax = m.plot(column=np.log10(m.units.fillna(0) + 1), cmap="YlOrRd", missing_kwds={"color": "lightgrey"}, figsize=(10, 6))
ax.set_axis_off()
counties drawn: 3,109; with a permit row: 2,995; 1,000+ units: 329
Map of United States counties shaded by housing units permitted in 2025, darkest around Phoenix, Texas cities, Florida and the Southeast, palest across the Great Plains, with grey counties that have no permit row. Map of United States counties shaded by housing units permitted in 2025, darkest around Phoenix, Texas cities, Florida and the Southeast, palest across the Great Plains, with grey counties that have no permit row.
Housing units permitted by county in 2025; grey counties have no permit-issuing place.

Of the 3,109 counties drawn in the lower 48, 2,995 have a permit row and 329 permitted 1,000 or more units. The grey counties are the ones with no permit-issuing place.

FAQ

Where do I download US building permit data by county?

From the Census Bureau's BPS county folder: coYYYYa.txt is the annual file for year YYYY, coYYMMy.txt the year to date to month MM. They are comma-separated text with two header rows.