↖︎ Vishal Singh
Contents

Zillow housing data · Chapter 2 of 5

Zillow housing data · Chapter 2 of 5

The geography ladder, from states to neighborhoods

Zillow estimates each level of geography on its own, so a state's typical home value differs from the average of its counties' values; in October 2026 New York State's was 19.6% below the housing-unit-weighted mean of its 62 counties, the largest gap of any state, while half of the states were between 2.8% below and 0.7% above theirs.

Author
AffiliationNYU Stern
Published

Zillow, November 2026 release

Zillow estimates each level of geography on its own, so a state's typical home value differs from the average of its counties' values; in October 2026 New York State's was 19.6% below the housing-unit-weighted mean of its 62 counties, the largest gap of any state, while half of the states were between 2.8% below and 0.7% above theirs.

Author

Vishal Singh

Affiliation

NYU Stern

Published

November 19, 2026

Data

Zillow, November 2026 release

Zillow's November 2026 release publishes a typical home value for the United States and for 51 states (the 50 states and the District of Columbia), 894 metro areas, 3,107 counties, 24,197 cities, 27,996 ZIP codes and 23,568 neighborhoods [1]. Each series is estimated from the Zestimates of the homes inside that region [2]. This chapter describes how much of the country each level covers and since when, which levels fit inside others, how far a published state value sits from the average of its counties' values, and how the differences in 10-year growth across ZIP codes divide between states, metros, counties and ZIP codes.

Counties with a series held 73% of U.S. housing units in March 2000 and 99.97% in October 2026

Table 1 lists the six levels. The smoothed files start in March 2000, and at every level below the state only part of today's series reach back that far. In March 2000, 47 states had a value (Montana, New Mexico, North Dakota and Wyoming start later; all 51 have values from March 2009), as did 430 of the 894 metros (48.1%), 1,034 counties (33.3%), 8,696 cities (35.9%), 12,892 ZIP codes (46.0%) and 11,255 neighborhoods (47.8%). The counts of county, city, ZIP-code and neighborhood series rose through the whole period (Figure 1). A few series also skip months after they start: 1,666 ZIP codes, 424 counties, 165 metros and 6 states have at least one blank month between their first and last values.

A ZIP code is a set of mail delivery routes assigned by the U.S. Postal Service, so it has no official boundary. The Census Bureau approximates ZIP codes with ZIP Code Tabulation Areas (ZCTAs), areas built from census blocks, and publishes its tables for them [3]. Here each Zillow ZIP code is matched to the ZCTA with the same five digits; 27,980 of the 27,996 have one.

Counting homes gives higher coverage than counting series. The American Community Survey (ACS) counts 143.78 million housing units in the 50 states and the District of Columbia over 2020 to 2024 [4]. Counties with a series held 73.4% of them in March 2000 and 99.97% in October 2026. The ZCTAs matched to Zillow ZIP codes held 71.7% and 99.3%, and the counties assigned to a metro with a series 77.7% and 92.5%. The series that start later are, on average, in places with fewer homes. City and neighborhood boundaries do not match Census areas one to one, so their housing coverage is not computed here.

Table 1. Counties and ZIP codes with a series held 99.3% or more of U.S. housing units in October 2026, against 71.7% to 73.4% in March 2000
Series in Zillow's typical home value files (all homes, mid tier, smoothed and seasonally adjusted), November 2026 release
LevelSeries in fileWith a value, March 2000All series have values fromU.S. housing units covered, March 2000Same, October 2026Location columns in the file
United States11March 2000100%100%RegionName
States5147 (92%)March 200998.5%100%RegionName (full state name)
Metros894430 (48%)June 201977.7%92.5%RegionName, StateName (first state of the metro name)
Counties3,1071,034 (33%)April 202673.4%99.97%State, Metro, StateCodeFIPS, MunicipalCodeFIPS
Cities24,1978,696 (36%)October 2026––State, Metro, CountyName
ZIP codes27,99612,892 (46%)October 202671.7%99.3%State, City, Metro, CountyName
Neighborhoods23,56811,255 (48%)October 2026––State, City, Metro, CountyName
Table 1. The finer the level, the smaller the share of today's series that reach back to 2000. Housing units covered: share of U.S. housing units in states, in counties assigned to a metro with a series, in counties, and in ZCTAs matched to ZIP codes, counting a region once its series has started. Dashes mark levels whose boundaries do not match Census areas.
Source: Zillow Home Value Index files, November 2026 release [1]; U.S. Census Bureau, American Community Survey 2020-2024 5-year estimates, table B25001 [4]. Housing units are 2020-2024 averages applied to every month. Counts unweighted.
The number of county, city, ZIP-code and neighborhood series grew 2.1 to 3.0 times between March 2000 and October 2026
Figure 1. Every line rises; series enter the files over time, and every series in the November 2026 files has a value in October 2026. Each line counts the series at one level with a value in each month, March 2000 to October 2026 (vertical axis on a log scale). The second view shows the share of U.S. housing units in the areas with a value, for the four levels that match Census areas. Hover for monthly values.
Source: Zillow Home Value Index, all homes, mid tier, smoothed and seasonally adjusted, November 2026 release [1]; ACS 2020-2024 5-year estimates, table B25001 [4]. Counts are unweighted; housing shares use 2020-2024 housing units for every month.

Counties fit inside states and metros; metros and ZIP codes cross county and state lines

Every county in Zillow's county file has one state, and its FIPS code agrees with that state. The file's Metro column gives each county at most one metro: 1,836 counties carry a metro name, and 1,271, with 6.4% of U.S. housing units, carry none. The column holds 927 metro names. Of these, 894 match a series in the metro file; the other 33, covering 1.0% of housing units, have no metro series under a matching name. The largest are Dayton-Kettering, OH, Poughkeepsie-Newburgh-Middletown, NY and Prescott Valley-Prescott, AZ. The Poughkeepsie name is the one Zillow gives to Dutchess and Orange counties, north of New York City, so those two counties belong to no metro series.

Metros do not fit inside states. Of the 927 metro names, 61 have counties in more than one state; 9 span three or more, and 2 span four (Philadelphia and Washington). These 61 metros hold 23.8% of U.S. housing units. The New York metro (Zillow's "New York, NY", the New York-Newark-Jersey City metro) is one of them. Its 23 counties are 10 in New York, 12 in New Jersey and 1 in Pennsylvania (Pike County), with 5.28 million, 2.76 million and 40,200 housing units. Its 848 ZIP code series are 448 in New York, 386 in New Jersey and 14 in Pennsylvania. The metro holds 61.5% of New York State's housing units and 72.8% of New Jersey's, so the state and metro series overlap only in part.

ZIP codes cross county lines. Zillow's ZIP file lists one county for each ZIP code. The Census Bureau's 2020 ZCTA-to-county relationship file records the land area of each piece of a ZCTA that falls in each county [5]. Of the 33,659 ZCTAs in the 50 states and the District of Columbia, 10,106 (30.0%) have land in more than one county, and 6,421 (19.1%) have less than 90% of their land in any single county. These hold 25.6% and 13.3% of ZCTA housing units. Of all ZCTAs, 137 cross a state line. For the 9,416 Zillow ZIP codes whose ZCTA has land in more than one county, the county Zillow lists is the one holding the largest share of the ZCTA's land in 91.0% of cases. The relationship file has no housing counts for the pieces, so it cannot show how many homes in a ZIP code lie outside the county Zillow assigns.

Zillow's ZIP file records a ZIP code's metro separately from its county. For 1,017 ZIP codes, the metro in the ZIP file differs from the metro of the ZIP code's own county in the county file; in 1,016 of these, the county has no metro at all.

Cities and neighborhoods also carry one state, one metro and one county each. The city file assigns New York City to Queens County, although the city spans five counties. The neighborhood file adds a City column: its 23,568 neighborhoods lie in 1,375 cities, 5.7% of the cities with a series, and for 99.9% of neighborhoods that city appears in the city file. Nashville has the most neighborhoods (417); New York City has 228. In 74 cities the neighborhoods carry more than one county; New York City's neighborhoods carry all 5 of its counties. The neighborhood file has no ZIP code column, so Zillow's files do not place neighborhoods in ZIP codes. Chapter 5 matches New York City's neighborhoods to Census areas.

Table 2. Counties nest in states and metros, while 61 of 927 metros cross a state line and 30.0% of ZCTAs cross a county line
How each level relates to the levels above it in Zillow's files and in Census geography
Smaller unitLarger unitWhat Zillow's files recordWhat the boundaries show
CountyStateOne state per countyAll 3,107 counties lie in one state
CountyMetroAt most one metro per county1,836 counties in a metro, 1,271 outside; 33 metro names have no metro series
MetroStateNothing (the metro name lists its states)61 of 927 metros span two to four states, with 23.8% of U.S. housing units
ZIP codeCountyOne county per ZIP code30.0% of ZCTAs have land in more than one county
ZIP codeMetroOne metro per ZIP code, set apart from its county's1,017 ZIP codes carry a metro different from their county's
ZIP codeStateOne state per ZIP code137 ZCTAs cross a state line
CityCountyOne county per cityNew York City is listed in Queens County
NeighborhoodCityOne city per neighborhood99.9% name a city in the city file
NeighborhoodZIP codeNothingNot linked in Zillow's files
Table 2. Only the county-in-state and county-in-metro relations hold without exception. Each row compares the single location Zillow records for a region with what Census boundaries show. ZCTA counts use land area from the 2020 relationship file.
Source: Zillow Home Value Index metadata, November 2026 release [1]; U.S. Census Bureau, 2020 ZCTA-to-county relationship file [5]; ACS 2020-2024 5-year estimates, table B25001 [4]. Counts of regions are unweighted.

Most state values sit close to the housing-weighted average of their counties; New York's sits 20% below

Zillow computes each region's value from the Zestimates of the homes inside it: each month it takes a trimmed mean of the middle third of those Zestimates by value and chains it to an index of month-to-month changes in the same homes' Zestimates [2]. A state's value is computed from the state's own homes, so it need not equal any average of its counties' values. To measure how far apart they are, each state's October 2026 value is compared here with the mean of its counties' values, weighted by each county's housing units in ACS table B25001 [4].

The state value was below its counties' weighted mean in 36 states and above it in 15. Half of the states were between 2.8% below and 0.7% above, and 5 differed by more than 5% (Figure 2). The District of Columbia, a single county, serves as a check; its state and county series differ by 0.003%. Maine had the largest positive gap (+4.3%). New York had the largest negative gap, followed by Wyoming (−17.5%). Across the 50 states with more than one county, the gap was more negative where county values were more spread out: the correlation between the gap and the housing-weighted standard deviation of log county values was −0.51.

New York State's value was $523,200, and its 62 counties' weighted mean was $650,700. Only 9 counties had values above the state's, but they hold 55.0% of the state's housing units. The five counties of New York City hold 42.8%; their weighted mean was $904,300, against $460,700 for the other 57 counties. County values ranged from $142,200 in Allegany County to $1.26 million in New York County (Manhattan).

Ten-year changes show a spread of similar width. From October 2016 to October 2026, the state's change minus its counties' weighted mean change was between −1.5 and +2.2 percentage points for half of the states, and between −4.9 and +11.4 for every state except New York. Colorado had the most negative gap. New York's gap was the largest. The state series rose 80.7% and its counties' weighted mean 59.1%, a gap of 21.6 percentage points. The state's change was closer to that of the 57 counties outside New York City (weighted mean 84.8%) than to that of the five boroughs (24.9%).

Table 3 repeats the comparison for the New York metro and the United States. The New York metro's value was $730,100, 8.2% below the weighted mean of its 23 counties ($795,400), and its 10-year rise of 66.7% was 14.2 percentage points above theirs. The U.S. value of $364,300 was 9.6% below the housing-weighted mean of the states, 14.4% below that of the counties and 19.2% below that of the ZIP codes. The U.S. 10-year change (76.6%) was within 2.0 percentage points of the weighted mean change at each level. The comparison describes how far the published series sit from a weighted average of their parts; it does not identify which step in Zillow's method produces the gap.

In October 2026 half of the states' typical home values were within 2.8% of their counties' weighted mean; New York's was 20% below, and its 10-year rise 22 percentage points above theirs
Figure 2. New York sits at the end of both views, below its counties in level and above them in 10-year change. Each bar is one state's published value minus the housing-unit-weighted mean of its counties' values, as a percent of that mean (level) or in percentage points (10-year change, October 2016 to October 2026). New York is drawn in orange and labeled in bold. Hover for the two values.
Source: Zillow Home Value Index, all homes, mid tier, smoothed and seasonally adjusted, state and county files, November 2026 release [1]. Weights: ACS 2020-2024 5-year estimates, table B25001 [4]. Counties with a value in the month(s) compared; they hold at least 86.8% of each state's housing units (North Dakota is lowest).
Table 3. Published values sit 8.2% to 19.6% below the housing-weighted means of their parts; U.S. 10-year changes sit within 2.0 percentage points of theirs
Published typical home value and 10-year change compared with the housing-unit-weighted mean of the regions one level down
ComparisonPublished, Oct. 2026Weighted mean of partsGapPublished change, 2016-2026Weighted mean of parts' changesGapHousing units covered
United States vs. 51 states$364,300$402,900−9.6%76.6%76.5%+0.1 pp100%
United States vs. 894 metros$364,300$433,800−16.0%76.6%75.1%+1.5 pp92.5%
United States vs. 3,107 counties$364,300$425,500−14.4%76.6%74.7%+2.0 pp99.97%
United States vs. 27,980 ZIP codes$364,300$450,900−19.2%76.6%78.3%−1.6 pp99.3%
New York State vs. its 62 counties$523,200$650,700−19.6%80.7%59.1%+21.6 pp100%
New York metro vs. its 23 counties$730,100$795,400−8.2%66.7%52.6%+14.2 pp100%
Table 3. Every published value is below the weighted mean of its parts, by 8.2% to 19.6%. Gap in level: published value divided by the weighted mean, minus one. Gap in change: published percent change minus the weighted mean of the parts' percent changes, in percentage points (pp). Changes use parts with values in both October 2016 and October 2026 (2,970 counties and 23,327 ZIP codes for the U.S.).
Source: Zillow Home Value Index, all homes, mid tier, smoothed and seasonally adjusted, November 2026 release [1]. Weights: ACS 2020-2024 5-year estimates, table B25001, by state, county or matched ZCTA [4]; metro housing units are the sum of the counties Zillow assigns to each metro.

About 30% of the variation in 10-year growth across ZIP codes lies between states and 33% within counties

Across the 23,336 ZIP codes with values in both October 2016 and October 2026, the 10-year rise in typical home value was 42.9% at the 10th percentile and 117.6% at the 90th. A nested variance decomposition splits these differences into parts that add up to the whole. Growth is measured as the log change (the natural log of the October 2026 value divided by the October 2016 value). Each ZIP code's distance from the national average is split into four steps: from the national average to its state's average, from the state's average to its metro's average within that state, from the metro's average to its county's average, and from the county's average to the ZIP code itself. Squaring each step and summing over ZIP codes gives four sums of squares that add up to the total. Metros that cross state lines are split by state, and ZIP codes with no metro form an "outside metros" group in each state. Averages and sums are weighted by the housing units of the matched ZCTA; the unweighted version counts each ZIP code once.

With housing-unit weights, 29.8% of the variation in 10-year growth lay between states, 23.4% between metros within a state, 13.6% between counties within a metro and 33.1% between ZIP codes within a county (Figure 3). Without weights the four shares were 30.3%, 22.9%, 17.6% and 29.2%. For the log level in October 2026, more of the variation lay between states (44.2% weighted) and less between counties within a metro (7.2%); the share within counties was 24.1%.

Some between-group variation would appear even if growth had nothing to do with location, because many groups are small: 995 county groups hold a single ZIP code. As a benchmark, the growth values were shuffled at random among the ZIP codes of the level above (nationally for states, within states for metros, within metros for counties), 50 times for each level. The shuffled shares averaged 0.5% for states, 5.0% for metros within states and 4.2% for counties within metros, below the shares in the data at every level.

The decomposition describes where the differences in growth sat between October 2016 and October 2026. It does not show that a ZIP code's state, metro or county produced its growth. The shares would change with a different period, different weights or different groups, and the groups here are Zillow's own metro and county labels, which, as described above, are assignments that do not always match Census boundaries. ZIP codes without an October 2016 value (4,660 of the 27,996) are left out.

Differences between states account for about 30% of the spread in ZIP-code growth from 2016 to 2026 and 44% of the spread in 2026 levels
Share of the variance of log 10-year change and of log level across ZIP codes at each level of nesting, percent
Figure 3. Between-state differences are the largest single part of the spread in levels; for growth with housing-unit weights, the within-county part is the largest. Each bar splits the total variance across ZIP codes into four parts that sum to 100%, ordered left to right from state to ZIP code. Growth is the log change from October 2016 to October 2026; level is the log of the October 2026 value. Weighted bars use ZCTA housing units (23,327 ZIP codes); unweighted bars count each of the 23,336 ZIP codes once.
Source: Zillow Home Value Index by ZIP code, all homes, mid tier, smoothed and seasonally adjusted, November 2026 release [1]. Weights: ACS 2020-2024 5-year estimates, table B25001, ZCTA with the same code [4]. Groups: 51 states, 1,040 state-metro groups (including 45 "outside metros" groups) and 3,529 county groups, from Zillow's ZIP file.
Numbers behind Figure 3
Part of the varianceGrowth, weightedGrowth, unweightedLevel, weightedLevel, unweightedGrowth, shuffled benchmark
State29.8%30.3%44.2%39.9%0.5%
Metro within state23.4%22.9%24.5%28.6%5.0%
County within metro13.6%17.6%7.2%9.4%4.2%
ZIP within county33.1%29.2%24.1%22.2%–

Exercise

Pick a state other than New York. From the November 2026 release, take the state's typical home value and its counties' values for October 2016 and October 2026, and the counties' housing units from ACS table B25001 (match on the five-digit county FIPS code; Connecticut needs the town crosswalk described in the appendix). Compute the housing-unit-weighted mean of the county values and of the county 10-year changes, and compare them with the state's own series. Then repeat with every county weighted equally. Report where your state falls in Figure 2, which counties have values above the state's, and whether equal weights move the gap toward zero or away from it.

What these data cannot show

Zillow publishes no margin of error for its series, so the gaps in Figure 2 and Table 3 cannot be set against the uncertainty of each estimate. The housing-unit weights are a 2020 to 2024 survey average applied to both 2016 and 2026, and they count rental apartments that the all-homes ZHVI leaves out. The relationship file describes 2020 ZCTA boundaries by land area; it does not show where homes sit inside a ZCTA. ZCTAs reflect ZIP codes as of January 2020 [3], and 16 Zillow ZIP codes have no ZCTA. Zillow's files do not state which metro delineation their labels follow. All dollar figures are nominal.

Data sources

  1. Zillow Research data files, November 2026 release (server date November 16, 2026), downloaded November 19, 2026 from files.zillowstatic.com/research/public_csvs/: the files State_, Metro_, County_, City_, Zip_ and Neighborhood_zhvi_uc_sfrcondo_tier_0.33_0.67_sm_sa_month.csv. Each file's URL, size, SHA-256 and server date: booklet/sources/manifest_2026-11-19.csv.
  2. U.S. Census Bureau, American Community Survey 2020-2024 5-year estimates, table B25001 (housing units), for the nation, states, counties, Connecticut county subdivisions and ZCTAs. Cached as zillow_download_cache/census/acsdt5y2024-b25001.dat.
  3. U.S. Census Bureau, 2020 ZCTA to county relationship file, https://www2.census.gov/geo/docs/maps-data/data/rel2020/zcta520/tab20_zcta520_county20_natl.txt, downloaded November 19, 2026. Pipe-delimited; columns used: GEOID_ZCTA5_20, GEOID_COUNTY_20, AREALAND_ZCTA5_20, AREALAND_PART. The file has land and water areas for each piece and no housing or population counts.
  4. U.S. Census Bureau, Connecticut county to county subdivision crosswalk, https://www2.census.gov/geo/docs/reference/ct_change/ct_cou_to_cousub_crosswalk.txt, downloaded November 19, 2026.

Methods

Coverage. A series counts as covered in a month if Zillow's file has a value for it. Housing-unit coverage counts a region from its first month with a value, using ACS 2020-2024 housing units for every month. The U.S. total (143.78 million) excludes Puerto Rico, which has no Zillow series. ZCTA rows in the ACS extract carry the prefix 860Z200US. Zillow ZIP codes are matched to ZCTAs on the five-digit code (16 have no match); counties on the five-digit FIPS code. Metro housing units are the sum over the counties whose Metro column names that metro; county metro names are matched to the metro file by first principal city and first state.

Connecticut. Zillow uses Connecticut's 8 former counties; the 2020-2024 ACS uses 9 planning regions. Housing units for the former counties are the sum of the ACS county subdivisions (towns) assigned to each in the Census crosswalk; the towns sum to the ACS state total (1,541,822 housing units).

Nesting. Multi-state metros are metro names in the county file whose counties carry more than one state code; in every case the count agrees with the states listed in the metro name. ZCTA splits use pieces with positive land area. A ZIP code's county is matched to the relationship file through the county file's FIPS codes (7 ZIP codes whose county name is not in the county file are left out).

States against counties. For each state, the housing-unit-weighted mean of its counties' October 2026 values, and the weighted mean of the counties' percent changes from October 2016 to October 2026, using counties with values in the months compared. Gap in level: state value over weighted mean, minus one. Gap in change: state percent change minus the weighted mean change. The correlation in the text uses the 50 states with more than one county.

Variance decomposition. For ZIP code i with weight wi and log change yi, let ȳ be the weighted national mean and ȳs, ȳsm, ȳsmc the weighted means of its state, state-metro group and county within that group. The total Σwi(yi − ȳ)² equals the sum of Σwi(ȳs − ȳ)², Σwi(ȳsm − ȳs)², Σwi(ȳsmc − ȳsm)² and Σwi(yi − ȳsmc)². Groups use the State, Metro and CountyName columns of the ZIP file. The weighted sample drops 9 ZIP codes whose ZCTA is missing or has no housing units. The level decomposition uses the same ZIP codes; with all 27,977 weighted ZIP codes that have an October 2026 value, the state share is 44.5%. Grouping counties by the county file's metro instead of the ZIP file's changes the weighted growth shares by at most 0.2 percentage points. Benchmark: for each level, the (growth, weight) pairs were shuffled among the ZIP codes of the level above, 50 times (seed 20261119), and that level's share recorded.

How to cite

@misc{singh2026zillowch2,
  author = {Singh, Vishal},
  title  = {The geography ladder, from states to neighborhoods},
  year   = {2026},
  note   = {In: Zillow housing data. NYU Stern. Data: Zillow Research},
  url    = {https://vishalsingh.org}
}

Acknowledgments

Home-value, rent and market estimates are from Zillow Research. Zillow requires clear attribution when its data are used.

References

  1. Zillow Research. Housing data. https://www.zillow.com/research/data/ (accessed November 19, 2026).
  2. Olsen, S. Zillow Home Value Index methodology, 2023 revision: what's changed? Zillow Research, February 10, 2023. https://www.zillow.com/research/methodology-neural-zhvi-32128/
  3. U.S. Census Bureau. ZIP Code Tabulation Areas (ZCTAs). https://www.census.gov/programs-surveys/geography/guidance/geo-areas/zctas.html (accessed November 19, 2026).
  4. U.S. Census Bureau. American Community Survey 2020-2024 5-year estimates, table B25001: Housing units. https://data.census.gov/table/ACSDT5Y2024.B25001
  5. U.S. Census Bureau. 2020 ZCTA to county relationship file. https://www2.census.gov/geo/docs/maps-data/data/rel2020/zcta520/tab20_zcta520_county20_natl.txt (accessed November 19, 2026).