The State of U.S. Parcel Records: Openness, Quality, and Gaps

Ask for parcel data by state and the answer arrives as a sales page. Coverage is nationwide, the boundaries are standardized, the file refreshes daily. None of that is false, and none of it answers the question a lender actually has, which is whether the record behind a specific site can be obtained, from whom, at what vintage, with which fields populated, and under terms that allow it to sit in a credit file. That question has a public answer, and it is now better documented than at any point in the history of American land records.

The documentation comes from the National States Geographic Information Council, whose 2025 Geospatial Maturity Assessment was published on August 14, 2026 from results collected in the second half of 2025. It graded the cadastre theme across all fifty states and the District of Columbia, county by county and municipality by municipality. Three of its headline numbers set up everything that follows: 73% of states (36 states and the District of Columbia) run a statewide parcel map program, 82% (41 states and the District of Columbia) have 100% coverage of digital parcel data, and 43% (21 states and the District of Columbia) have 100% publicly available parcel data through a feature API or a download (National States Geographic Information Council, 2025 Geospatial Maturity Assessment, 2026).

Read those three figures next to each other and the counterintuitive result falls out. The binding constraint on national parcel analytics is no longer whether the geometry exists. It exists almost everywhere. The constraint is whether the geometry can be obtained and whether its attributes follow any standard, and on both of those measures the country has barely moved in two years. Eight states run a statewide parcel program, hold complete or near complete digital parcel coverage, and still publish under a quarter of their jurisdictions: Georgia, whose 159 counties are all digital and none downloadable, Kansas with 105, Nebraska with 93, South Dakota with 66, Oregon with 36, New Mexico with 33, plus Michigan at 5% and South Carolina at 13%. Across the thirty-seven programs, which the survey counts as thirty-six states plus the District of Columbia, 764 counties or municipalities hold digital parcels that the public cannot download or query. And only 20 of those thirty-seven apply a parcel standard that includes verification and quality control.

What follows is a method rather than a directory. It covers who makes a parcel record and why there is no national one, what openness means once licences and fees are included, where quality actually breaks (attributes, geometry, vintage, identifiers), which gaps no coverage percentage will show you, and how to evaluate a parcel source before it enters a lending decision. It sits inside the wider argument about building open analytical atlases from public data. Readers weighing where to get parcel data should treat this piece as the diagnostic that comes first, and the companion work on parcel-derived land metrics as what becomes possible once the source survives the diagnostic.

What a parcel record is, and who makes it

A parcel record is not one document. It is at least three, produced by different offices on different cycles under different law, and the habit of speaking about them as a single thing causes most of the confusion in this field.

The first is the assessment roll, maintained by an assessor or an appraisal district for the purpose of levying property tax. It carries the parcel identifier, the owner of record for tax billing, a situs address, a land use or property class code, assessed and sometimes market values, land area, and in many jurisdictions a set of building attributes. The second is the recorded instrument, held by a county recorder, register of deeds or clerk, which carries the legal chain of title: deeds, mortgages, liens, easements, plats. The third is the parcel polygon, a GIS representation of the assessment unit, usually drawn by a county land information office or mapping department from plats, deeds and survey control. Only the third is a map. The first two are ledgers, and the polygon is joined to the roll by a key that neither office is obliged to keep stable.

The number of offices producing these records is larger than the familiar figure of roughly 3,100 counties suggests. The 2022 Census of Governments counted 90,837 local governments in the United States, of which 38,736 are general purpose: 3,031 county governments, 19,491 municipal governments and 16,214 township governments, with 39,555 special districts and 12,546 independent school districts making up the rest (U.S. Census Bureau, 2022 Census of Governments, Organization component, 2023). Township governments exist in exactly twenty states, led by Minnesota with 1,779, Pennsylvania with 1,546, Illinois with 1,425, Ohio with 1,308 and North Dakota with 1,305. In New England and parts of the Mid-Atlantic the assessing unit is the town or municipality rather than the county, which is why the NSGIC survey counts Maine's producing jurisdictions at 530 rather than at its sixteen counties.

That distinction is not academic for an analyst. It determines how many separate acquisitions, licences, schemas and refresh cycles stand between a question and an answer. Texas has 254 counties to reach through their appraisal districts, Georgia 159 counties, Virginia 133, Kentucky 120, Missouri 115. At the other end the District of Columbia is one jurisdiction, Delaware is three, Hawaii is five and Connecticut is nine. A national parcel layer is the sum of every one of these negotiations, repeated annually.

The practical consequence is that jurisdiction count, not state count, is the right denominator for most coverage questions. A claim that a source covers forty-eight states can be true while a third of the country's producing offices are absent from it, and the two statements are not in conflict.

Why the United States has no national cadastre

Most countries with a cadastre built it as an instrument of national taxation or national land registration, administered centrally. The United States did neither. Property taxation is local, land registration is a state matter conducted through recording rather than registration of title, and the federal survey system was designed to dispose of public land rather than to maintain a permanent record of private holdings. The result is a country with an extremely detailed land record and no national copy of it.

There is a federal coordination layer, and it is worth understanding precisely because its limits are so often misread. OMB Circular A-16, revised August 19, 2002, establishes the coordinated approach to developing the National Spatial Data Infrastructure and establishes the Federal Geographic Data Committee, directing federal agencies that produce, maintain or use spatial data (Office of Management and Budget, Circular A-16, 2002). The Geospatial Data Act of 2018 was signed on October 5, 2018 as a component of the FAA Reauthorization Act, H.R. 302, Public Law 115-254, and is codified at 43 U.S.C. Chapter 46, with section 2802 covering the Federal Geographic Data Committee, section 2804 the National Spatial Data Infrastructure, section 2805 the National Geospatial Data Asset data themes, section 2806 geospatial data standards and section 2808 covered agency responsibilities (Federal Geographic Data Committee, Geospatial Data Act of 2018, 2026). Cadastre is one of the National Geospatial Data Asset themes; the original list of 34 NSDI data themes was revised to 16 NGDA themes on August 19, 2011 and the official theme list was updated on July 3, 2024 to incorporate the Act's mandates.

Read the chain and the gap is obvious. Every one of those instruments binds federal agencies. None of them binds a county assessor. The Cadastral Subcommittee of the FGDC, led by the Bureau of Land Management, "develops and implements plans to coordinate cadastral data-related activities among Federal, State, Tribal, and local governments, and the private sector", and works with the public domain states, meaning those covered by the Public Land Survey System, "to standardize the best available statewide data in the State, to identify the optimum maintenance work flow, and to publish a maintainable standardized dataset" (Federal Geographic Data Committee, Cadastral Subcommittee, 2026). Coordination, standardization and publication assistance are the verbs. Compulsion is not among them.

The federal contribution that does exist at national scale is the survey framework rather than the parcels. The BLM facilitated or completed standardization of 29 of the 30 PLSS states, with Oklahoma pending at the time of the subcommittee's reporting, and the standardized PLSS is itself a National Geospatial Data Asset (Federal Geographic Data Committee, Cadastral Subcommittee, 2026). That matters for geometry, because in the thirty public land survey states a parcel can be tied to a section, township and range framework with published control, while in the colonial and state land states the description is metes and bounds and the framework is whatever local surveyors and plat filings have produced. The same subcommittee has coordinated the construction of geodatabases for over 500,000 Indian parcels on reservations identified for the Land Buy-Back Program for Tribal Nations, using standard PLSS data combined with land descriptions from the Bureau of Indian Affairs trust asset system, which is a reminder that tribal land is a separate records regime rather than a hole in the county roll.

Openness by state: what the 2025 survey found

The 2025 assessment is the first cadastre survey to grade every state and the District of Columbia, and the first to publish sub-grades for each component of the overall grade. It measures four things: whether a statewide parcel map program exists, the share of counties or municipalities whose parcels exist in digital form, the share whose parcels are publicly accessible through a download or a feature API, and the standard the program applies to what it receives. Program characteristics (a designated steward, funding, a business plan, formal relationships with local government, and attribute collection) carry a fifth sub-grade.

Twenty states and the District of Columbia sit at exactly 100% on both digital and public access in the state grade table, and the report's own summary statistic for complete public access counts 22, one more than the table shows at exactly 100%. Ranked by the number of jurisdictions each has to reach, that group runs from North Carolina at 100 counties, Indiana at 92, Ohio at 88, Mississippi at 82, Arkansas at 75, Wisconsin at 72, Florida at 67, Montana at 56, West Virginia at 55 and North Dakota at 53 down to the District of Columbia at one. Texas is a near miss at 253 of 254 counties, or 99.6%, and Tennessee publishes 86 of 95 counties (National States Geographic Information Council, 2025 Geospatial Maturity Assessment, 2026).

The interesting half of the table is the other one. In the states with the widest gap between what exists and what is published, digital coverage is complete and public access is close to zero. Georgia, Kansas, Nebraska, South Dakota and Oregon each report 100% digital parcel data and 0% public access. New Mexico reports 82% digital and 0% public. Michigan publishes 4 of 83 counties, South Carolina 6 of 46, Idaho 13 of 44, New York 36 of 62. These are not states without capacity. Georgia, Kansas, South Dakota and Oregon all earn an A on the digital access sub-grade, and Georgia, Kansas, South Dakota, Oregon and New Mexico all earn an A on program characteristics. They have the data and the institutions. What they do not have is a decision to publish.

The fourteen states without a statewide program deserve a separate reading, because the survey measures them differently. For them the published figure is the maximum share of counties or municipalities reachable through a parcel viewer, a download or an API. A viewer is not a dataset. Iowa, Oklahoma, Alabama and New Hampshire reach 100% on that weaker test, Virginia and California 98%, Rhode Island 95%, Louisiana 86%, Illinois 81%, Missouri 77%, Nevada 76%, Maine 71%, Alaska 29% and Kentucky 4%. Anyone comparing a state from this group with a state from the first group is comparing two different questions, and the survey says so explicitly in its own methodology.

Alaska carries a further wrinkle worth noting because it generalizes. Its 29% is measured on land area covered by parcels rather than on a straight share of boroughs and census areas, because some are not fully covered. Where parcel mapping is incomplete inside a jurisdiction, a jurisdiction count overstates coverage and an area count understates the share of value covered, since the unmapped parts are usually the least developed. Neither number is wrong. Both need their denominator printed next to them.

Two survey cycles, and where the needle did not move

A single survey is a snapshot. Two are a trend, and the trend is the strongest evidence for the argument of this piece. The 2023 Geospatial Maturity Assessment, published in December 2023, drew responses from 44 states on the cadastre theme. Of the 35 with parcel data aggregation programs, 22 made 80% to 100% of their parcels available to the public, 12 made them only partially available, and 5 kept parcel data for internal use only. A majority of programs collected parcel attribute data (91%), had a designated steward (85%) and had formal relationships with local government (69%). A slight majority had regular state funding (57%) and fewer than half had a business plan (47%). The nine states without programs relied mainly on county parcel viewers (National States Geographic Information Council, 2023 Geospatial Maturity Assessment, 2023).

Set the two cycles side by side on the measures that carry over. Program states publishing 90% to 100% of their jurisdictions went from 22 of 35 to 23 of 37. Program states publishing under 25% went from 9 to 8. Two years, one state. On the program characteristics, a designated steward went from 30 states to 33, regular funding from 20 to 23, a business plan from 15 to 16, formal relationships with local government from 24 to 26, and attribute collection from 32 to 34, each against a denominator that itself grew from 35 to 37 (MMCG tabulation from the NSGIC 2023 and 2025 state report cards).

The standards picture moved slightly further. Programs applying a standard with verification and quality control rose from 17 to 20, those applying a standard without verification fell from 10 to 9, and best effort and as received held at 5 and 3. That leaves 17 of 37 programs, nearly half, accepting county parcel data either on a best effort basis or exactly as received. A statewide layer assembled that way is a container, not a schema, and the difference shows up the moment an analyst tries to filter on a land use code across county lines.

One caution about comparing the cycles, and it is the same caution this article keeps returning to. The 2023 survey covered the 44 states that responded on the theme. The 2025 survey covered all fifty states and the District of Columbia, with grades for non-respondents assigned from research and quality control by the theme graders. Three states, Arizona, Mississippi and South Dakota, submitted for the first time under the programs category, which improved their grades. Some 2025 grades were reduced because a state had reported parcels as publicly available in particular jurisdictions that further evaluation reclassified as internal only. The direction of the noise is therefore toward more accuracy, not toward flattery, which strengthens rather than weakens the finding that the bottom of the distribution is sticky.

Licences, fees, and the difference between open and reachable

Openness in the survey sense means the data can be downloaded or queried without a fee. It does not mean the data carries an open licence, and for a lending file the licence matters as much as the download.

Three regimes are common. The first is an explicit public domain or attribution posture. Utah's statewide parcel layer states that "There are no constraints or warranties with regard to the use of this dataset" and asks users to attribute content to the State of Utah (Utah Geospatial Resource Center, Utah Parcels, 2026). The second is a fitness disclaimer with no licence at all, which is the most common. New York's program states that "GIS parcel data is for planning and general use purposes only and is not a substitute for a legal parcel survey" (New York State GIS Program Office, Statewide Parcel Data, 2026). Texas states that its parcels "are not survey grade and should not be used for legal purposes" (Texas Geographic Information Office, StratMap Land Parcels, 2026). Wisconsin states that "No warranty, expressed or implied, is made regarding accuracy, completeness, or legality of the information herein" (Wisconsin State Cartographer's Office and Wisconsin Land Information Program, Statewide Parcel Map Initiative, 2026). The third regime is a negotiated agreement with the county, which is where fee-based access and redistribution restrictions live.

None of these disclaimers is a reason to avoid the data. They are a reason to be precise about what a parcel polygon is evidence of. It is evidence of the assessment unit as the assessor understands it. It is not evidence of a boundary, and no statewide program in the country claims otherwise. A credit memo that cites a parcel polygon as the site boundary has overstated its source; one that cites it as the assessor's mapped extent of the tax parcel, with the survey question flagged for the title and survey work, has not. The same discipline applies to every layer in a screen, which is the argument made at length in the piece on why every number needs a source and a date.

Fee regimes deserve one specific warning. Where a county sells its parcel file, the terms frequently restrict redistribution rather than use, and the restriction runs with the data. An analyst who buys a county extract, joins it to a national frame and ships the result to a client has done something the licence may not permit, even though every input was lawfully obtained. This is not a hypothetical distinction for lenders, because a market study delivered into a loan file is a distribution. The safe posture is to read the terms before the join, not after.

Quality, part one: the attributes

Geometry gets the attention and attributes carry the analysis. A polygon tells you where; the roll tells you what, whose, how big, how old and how it is taxed. The fields that matter for commercial work are a short list: parcel identifier, situs address, owner name and mailing address, land use or property class code, land area, assessed and market values by land and improvement, sale date and sale price, and building attributes.

Completeness varies by field in a predictable order. Parcel identifier and land area are near universal. Situs address is common but frequently blank on unimproved land, which is precisely the land a development screen cares about, and this is the single most under-appreciated join failure in site work, a problem the piece on the Census geocoder against commercial services takes up in detail. Land use codes exist almost everywhere and mean almost nothing across county lines: one county's code 340 is another's 4200, and a statewide program that accepts data "as received" has not reconciled them. Assessed values are present but are a statutory construct rather than a market observation, subject to assessment ratios, caps, exemptions and reassessment cycles that differ by state and sometimes by property class within a state.

Where states have standardized, the field list is worth reading closely. Utah publishes two products: a basic parcel layer with parcel identifier, address, city, ZIP code, ownership type and recorder contact fields, and a Land Information Records layer that adds tax roll attributes including assessor source, boundary source, current as of date, serial number, tax exempt type, tax district, total market value, land market value, parcel acreage, property class, primary residence status, house count, subdivision name, building square footage, floors count, built year, effective built year and construction material, "when available" (Utah Geospatial Resource Center, Utah Parcels, 2026). That qualifier is doing real work. A standardized schema guarantees a column, not a value.

Building attributes are the deepest and least reliable part of the roll, and they are treated separately in the companion analysis of what assessor rolls record about buildings. The short version for this piece: year built, building area and construction class are collected by most assessors because they drive the cost approach to value, but their definitions are local, their update discipline follows permit activity rather than a schedule, and their absence is unmarked. A blank year built field can mean unimproved land, an unpermitted structure, or a record the assessor never had to complete.

Sale price is the field that fails hardest, and it fails by statute rather than by neglect. In several states the instrument that carries the consideration is confidential. Montana is explicit: the realty transfer certificate "and the information contained in the certificate are not a public record and must be held confidential by the county clerk and recorder and the department", on the stated legislative finding "that the demands of individual privacy outweigh the merits of public disclosure", with an exception for compilations, summaries and analyses derived from the certificates (Montana Code Annotated 2025, section 15-7-308). Kansas restricts the real estate sales validation questionnaire to a closed list of recipients: county officials, a property owner prosecuting an appeal, county appraisers, licensed appraisers, financial institutions conducting appraisals and evaluations required by regulators, appeal bodies, county commissioners, the director of property valuation and licensed brokers acting for clients (Kansas Statutes Annotated, section 79-1437f, 2026). Note what the Kansas list includes: a financial institution conducting an appraisal or evaluation under regulatory requirement has statutory access that a market analyst does not.

The wider point for method is that the count of so-called non-disclosure states, usually given as twelve in trade sources, is not a figure any federal or state primary source publishes, and the states on the list do not restrict the same things in the same way. Some withhold the price from the public record entirely, some collect it on a confidential affidavit and publish only derived ratios, and some allow county-level variation. The defensible procedure is to test the sale price field county by county in the actual extract and record the result, rather than to inherit a list.

Quality, part two: the geometry

Parcel polygons are drawn, not surveyed. That single sentence explains most geometric complaints. A county mapping office compiles a parcel fabric from recorded plats, deed calls, aerial imagery and whatever survey control exists, and fits it to a base. The result is dimensionally reasonable at the scale of a neighbourhood and unreliable at the scale of a setback.

Four failure modes recur. The first is absolute positional error, which is largest where control is thinnest, typically rural and mountainous areas and the interiors of large unsubdivided tracts. The second is relative error at edges, where two adjacent parcels drawn from different plats overlap or leave a sliver. The third is edge matching at county lines, where two independent fabrics meet and neither is authoritative; a statewide program that clips rather than reconciles will carry the mismatch forward, and this is exactly where a trade area that straddles a county boundary produces double counted or missing land. The fourth is the treatment of non parcel space: rights of way, water and unmapped areas are sometimes polygons with null identifiers, sometimes gaps, and an area calculation that does not test for this will silently disagree with the assessor's own acreage field.

The public land survey framework helps in the thirty PLSS states, because parcels can be tied back to sections and quarter sections with published corner positions, and the BLM has standardized the PLSS in 29 of those 30 states (Federal Geographic Data Committee, Cadastral Subcommittee, 2026). In metes and bounds states there is no equivalent skeleton, and the fabric is only as coherent as the plats behind it. This is one reason terrain and constraint layers should be intersected with parcels carefully rather than confidently, a point the companion piece on screening buildable land by slope develops for elevation data specifically.

None of this argues against using parcels. It argues for stating the tolerance. For a screen that asks whether a site is roughly four acres or roughly fourteen, drawn parcels are more than adequate. For a question about whether a proposed building envelope clears a setback, they are not, and the answer has to come from a survey. The line between those two questions is where an analyst earns their fee, and it is the same line that governs the estimation of buildable area from a parcel.

Vintage: a statewide layer is not one date

Ask when a statewide parcel layer was last updated and the honest answer is a distribution, not a date. Utah publishes that distribution openly, which makes it the cleanest worked example in the country. The Utah Geospatial Resource Center holds parcel boundaries for each of the 29 counties and updates them "as new data is provided from the data steward in each county", posting the last update for every county on the public page (Utah Geospatial Resource Center, Utah Parcels, 2026).

Read as of August 24, 2026, five counties were refreshed that month and seven the month before. At the other end, Sanpete County's basic parcels were last updated in October 2015, which is 130 months, and Rich County's in July 2024. Daggett, Kane and San Juan sat at 13 months, Wayne at 12, Duchesne at 11 and Millard at 8. Twenty-one of the 29 counties were refreshed at some point in 2026, six in 2025, one in 2024 and one in 2015 (MMCG tabulation from the UGRC Utah Parcels page, read August 24, 2026).

The tax roll attached to the geometry carries its own separate clock. The Land Information Records layer shows a last update year of 2025 for 23 counties, 2024 for five, and 2020 for Box Elder. So a single statewide layer, in a state that grades A on every cadastre sub-grade and publishes 100% of its counties, contains geometry that ranges over a decade and attributes that range over six years. Nothing about that is a failure of the program. It is the arithmetic of aggregating from independent stewards, and every statewide layer in the country has the same structure whether or not it publishes the dates.

The operational rule follows directly. A parcel extract used in an underwriting file should carry a per-record vintage, not a file-level one, and where the source does not provide it the analyst should say so rather than adopt the download date. Any platform that displays parcel values without a vintage next to them is asking the reader to assume a currency the data does not have; MMCG Analytics is built on federal, state and public sources with source and vintage provenance carried on displayed values, and parcels are among those public record layers, which is a design response to exactly this problem rather than a claim to have solved it upstream.

Five programs, read the way an analyst should read them

Abstractions about openness resolve quickly when five named programs are put side by side on the same four questions: how many jurisdictions does the program collect from, how often does it say it refreshes, how many of those jurisdictions are publicly accessible, and what does it say about fitness for use.

Texas is the largest aggregation problem in the country. The Texas Geographic Information Office, a division of the Texas Water Development Board, publishes StratMap Land Parcels collected from the appraisal districts of the state's 254 counties, states plainly that "Not all counties are available for download from TxGIO" and that it "will attempt to refresh annually from each appraisal district or their third party vendor", with the caveat that "the refresh rate will vary across the state". File names carry the vintage year and month, which is the right design (Texas Geographic Information Office, StratMap Land Parcels, 2026).

Wisconsin runs the most systematized program. The State Cartographer's Office and the Wisconsin Land Information Program publish the Statewide Parcel Map Initiative on an annual version cycle; V12 carries 3.56 million records totalling 1.6 GB, collected from the counties in the first half of 2026, with V13 tentatively scheduled for June 30, 2027 (Wisconsin State Cartographer's Office and Wisconsin Land Information Program, Statewide Parcel Map Initiative, 2026). A published version number and a scheduled successor are worth more to a repeatable analysis than a continuously updated layer with no version identity, because they make a result reproducible.

New York shows the gap between a program and its coverage. The GIS Program Office, which created the program in 2014, publishes standardized tax parcel data for 38 counties in a single common schema, updated "once a year", and provides parcel centroids for all 62 counties (New York State GIS Program Office, Statewide Parcel Data, 2026). The 2025 assessment credits New York with public access in 36 of 62 jurisdictions. Both numbers are correct and they measure different things, which is a useful reminder that a state's own count and a national survey's count will rarely match exactly.

Montana is the cadence outlier. The Montana State Library provides cadastral data "monthly for each Montana county in file geodatabase and shapefile data formats", alongside the Department of Revenue's Orion computer assisted mass appraisal data by county, an ArcGIS REST endpoint and a public viewer (Montana State Library, Montana Cadastral, 2026). Utah sits between the two models, refreshing its five largest counties monthly and the rural counties on a rotating schedule that runs from quarterly to annual, with Land Information Records parcels updated annually.

The pattern across all five is that the state programs are candid about their limits and the aggregate presentations built on top of them usually are not. Every one of these pages tells a reader what is missing. That candour is a feature to look for, and its absence in any source is the strongest single signal available at no cost.

Identifiers that do not hold still

The parcel identifier is the join key for everything, and it is not stable. Assessors renumber. A split creates children with new identifiers and retires the parent. A merger does the reverse. Some counties encode the identifier with geographic meaning, so that a change in a taxing district or a map book changes the number without any change on the ground. Formatting varies: dashes, leading zeros, book page block lot structures, and the same identifier written three different ways in the roll, the recorder's index and the GIS layer of a single county.

Three consequences bear on lending work. First, a time series keyed on parcel identifier will lose records that did not disappear, and the loss looks like a decline in whatever is being counted. Second, a comparison of assessed values across a reassessment or a parcel split is a comparison of different objects, and land assembly analysis, which is the most valuable thing parcels can do, is precisely the case where identifiers change most. Third, there is no national parcel key and no federal identifier that can substitute for one; unlike a census tract or a county FIPS code, a parcel identifier is meaningful only inside its own jurisdiction and only until that jurisdiction decides otherwise.

The workable practice is to treat the identifier as a within vintage key and geometry plus address as the cross vintage key, to keep the source county code alongside the identifier so that duplicates across jurisdictions never collide, and to record identifier changes as events rather than silently dropping them. Anyone joining zoning to parcels faces the same problem from the other direction, since zoning districts are drawn on their own base and change on their own schedule, which is the subject of the piece on zoning data by parcel.

The gaps, and the denominator problem

Some of what looks like missing parcel data is not missing. It is absent because no assessment unit exists. Federal land is the largest case: land held by the United States is not on a county tax roll, so a parcel layer that shows nothing across a national forest or a military reservation is correct rather than incomplete, and any coverage percentage computed on land area rather than on taxable parcels will read as a failure where there is none. Tribal land is a separate regime again, recorded through the Bureau of Indian Affairs trust asset system rather than the county roll; the FGDC Cadastral Subcommittee reports that the BLM completed geodatabase construction and land records validation for over 500,000 Indian parcels across reservations identified for the Land Buy-Back Program for Tribal Nations, built from standardized PLSS data combined with BIA land descriptions (Federal Geographic Data Committee, Cadastral Subcommittee, 2026).

Genuinely missing data has four other sources. Unmapped interiors, where a jurisdiction has digitized its subdivided land and left large tracts as index polygons, which is why Alaska's coverage is reported on land area rather than on a count of boroughs and census areas. Jurisdictions that hold data internally, which is the 764 counties and municipalities inside states that have a program. Attribute gaps inside otherwise complete geometry. And the difference between a parcel polygon layer and a full assessment roll, which is the least visible of the four: several statewide programs publish geometry with a compact attribute set and leave the full roll, including the sale and valuation history, to be requested from the county. A source that reports parcel coverage and a source that reports roll coverage are answering different questions, and a buyer who does not ask which one is on offer will find out during a deal.

Which brings the argument to its practical centre. Every coverage claim about American parcels is true or false only relative to a denominator, and there are at least four defensible ones: states, jurisdictions, taxable parcels and land area. The 2025 assessment gives 73% for programs, 82% for digital coverage and 43% for public access when the unit is the state. Count jurisdictions inside the thirty-seven programs and the same survey yields 2,231 counties and municipalities, of which 2,219 have digital parcel data and 1,467 are publicly accessible, or 65.8%. Neither answer is wrong. A source that quotes one without naming it has not told you anything you can audit.

The habit worth building is to demand the denominator in writing. It is the same discipline that governs supply metrics such as self-storage supply per capita and exposure metrics such as the share of commercial land in the floodplain, where the numerator is usually easy and the denominator decides the answer.

How to evaluate a parcel source for a lending file

Five tests, in the order that kills a bad source fastest.

Provenance. Ask which office produced each record and whether that office is named in the data. A statewide program that carries an assessor source field, as Utah's Land Information Records layer does, allows an analyst to trace a suspect value to its origin. A layer with no source field cannot be audited at all, and the honest response to a challenged figure becomes an apology rather than a citation. Where a source is an aggregation of aggregations, ask how many hops separate the record from the assessor.

Vintage, per record. Ask for the date each record was received from its steward, not the date the file was assembled. If the answer is a single date for the whole country, the source is either extraordinarily well funded or reporting its own build date as the data's age. Test it: pick three jurisdictions known to refresh on different cycles, look at a parcel that transacted last year, and see whether the owner has changed.

The attribute dictionary. Ask for the field list with definitions and null rates by jurisdiction. Field lists are easy to obtain and null rates are not, which is why they are the more revealing request. Land use code is the field to press on, because a standardized code across counties is either a real crosswalk that someone maintains or a relabelling that quietly loses information. Sale price is the second, for the statutory reasons above.

Geometry lineage. Ask what the polygons were compiled from, whether the fabric is edge matched at county lines, and what positional accuracy the producer claims. Every state program in this article publishes a fitness statement; expect the same from any source and treat its absence as an answer. Then test the acreage field against computed polygon area on a sample; systematic divergence is normal and its size is diagnostic.

Licence terms, before the join. Ask what may be redistributed, what must be attributed and whether any county in the extract carries terms different from the rest. The answer determines whether the output can be delivered to a client, which is the only question that matters once the analysis is complete. The broader vendor conversation, including how these tests apply to commercial products, belongs with the questions to ask any data vendor and the comparison of what free federal sources actually cover.

A source that passes all five is rare. A source that fails the first or the fifth should not enter a file at all, because the failure is not about accuracy but about whether the work can be defended or delivered.

What public-only coverage supports today

Public parcel data, taken as it stands in 2026, comfortably supports a defined set of commercial questions. Site identification and land area screening across the twenty states and the District of Columbia at full public access, plus the near misses in Texas and Tennessee, covers a large share of the country's transaction volume. Ownership pattern and assembly analysis works wherever the roll is published, subject to the identifier discipline described above. Land use composition within a trade area works where a real code crosswalk exists, and is unreliable where the state accepts data as received. Intersections with hazard, terrain and infrastructure layers work everywhere the geometry exists, because those layers are federal and national and the parcel is only supplying the footprint, which is the pattern behind a 30 minute pre-term-sheet site screen.

Three things public-only coverage does not support. A national valuation comparison, because assessed value is a statutory construct with different ratios, cycles and exemptions by state. A national sale price series, because of the confidentiality statutes and the county level variation inside them. And any claim of complete national coverage, because 764 jurisdictions inside program states publish nothing at all and fourteen states have no statewide program to ask.

The correct posture for a lending team is neither to dismiss public parcels as incomplete nor to treat a national layer as authoritative. It is to know which of the four denominators applies to the question in hand, to carry the vintage on every record, and to state the limit in the memo rather than let a reader assume it away. That is a lower standard than a cadastre and a considerably higher one than most parcel analysis currently meets, and it is achievable at no cost in most of the country today. The parallel argument about how mapping tools differ from analytics tools is made in the piece on GIS platforms against CRE analytics platforms.

Frequently asked questions

Which states have free statewide parcel data available to the public?

As of the 2025 Geospatial Maturity Assessment, twenty states and the District of Columbia have 100% of their counties or municipalities publicly accessible through a download or feature API: North Carolina, Indiana, Ohio, Mississippi, Arkansas, Wisconsin, Florida, Montana, West Virginia, North Dakota, Washington, Utah, Maryland, Wyoming, New Jersey, Massachusetts, Vermont, Connecticut, Hawaii, Delaware and the District of Columbia. Texas reaches 253 of 254 counties, or 99.6%, and Tennessee 86 of 95 counties. Thirty-six states and the District of Columbia run a statewide parcel program, thirty-seven in all, so a program is not the same thing as public access.

Is there a national parcel database for the United States?

Not from the federal government. Parcel records are produced by roughly 3,100 counties plus municipalities and townships in states where the assessing unit is sub-county, and no federal statute requires them to be shared. OMB Circular A-16 and the Geospatial Data Act of 2018 direct federal agencies, not local assessors. The FGDC Cadastral Subcommittee, led by the Bureau of Land Management, coordinates and helps standardize state datasets, and BLM has standardized the Public Land Survey System in 29 of the 30 PLSS states, but the parcels themselves stay local. National coverage is assembled by aggregation, state by state and county by county.

Why do some states have digital parcels that the public cannot download?

Because digitization and publication are separate decisions. Eight states with a statewide program report complete or near complete digital parcel coverage and publish under a quarter of it: Georgia, Kansas, Nebraska, South Dakota, Oregon and New Mexico publish none, Michigan publishes 4 of 83 counties and South Carolina 6 of 46. Reasons vary and include county revenue from data sales, agreements with local government that predate the state program, and privacy or liability concerns. Across the thirty-seven programs, 764 counties or municipalities hold digital parcels the public cannot obtain.

What is a non-disclosure state for real estate sale prices?

It is a state where the price paid for real property is not part of the public record, usually because the transfer document that carries it is confidential by statute. Montana provides that the realty transfer certificate and its contents are not a public record and must be held confidential, with an exception for compilations and analyses. Kansas restricts the real estate sales validation questionnaire to a defined list of recipients that includes county appraisers, licensed appraisers, appeal bodies and financial institutions conducting appraisals or evaluations required by regulators. The commonly quoted count of twelve such states is not published by any federal or state primary source, and the restrictions differ in kind, so the reliable method is to test the sale price field county by county in the actual data.

How current is statewide parcel data?

It is a distribution rather than a date. Utah publishes the last update for each of its 29 counties: read on August 24, 2026, five counties had been refreshed that month and Sanpete County's parcels were last updated in October 2015, a spread of 130 months inside one statewide layer that grades A on every cadastre measure. Refresh policy also varies by program: Montana provides cadastral data monthly for each county, Utah refreshes its five largest counties monthly and rural counties quarterly to annually, New York updates once a year, Wisconsin publishes an annual version series, and Texas attempts an annual refresh from each appraisal district with a rate that varies across the state.

Can I use parcel boundaries to determine a property line?

No. Parcel polygons are compiled from plats, deeds, imagery and available control, not surveyed, and every statewide program says so. New York states that its parcel data is for planning and general use only and is not a substitute for a legal parcel survey. Texas states that its parcels are not survey grade and should not be used for legal purposes. Wisconsin makes no warranty as to accuracy, completeness or legality. Parcels are reliable for site identification, approximate land area, ownership pattern and layer intersection. Boundary questions require a survey.

What should I ask for before buying or using a parcel dataset?

Five things. The producing office for each record, named in the data. The date each record was received from that office, per record rather than per file. The field list with definitions and null rates by jurisdiction, with land use code and sale price examined first. The geometry lineage, including what the polygons were compiled from, whether the fabric is edge matched at county lines, and the positional accuracy claimed. And the licence terms, specifically what may be redistributed, since a market study delivered into a loan file is a distribution. A source that cannot answer the first or the last should not enter the file.

Sources

1. National States Geographic Information Council, 2025 Geospatial Maturity Assessment, Full Report, published August 14, 2026 from results collected in the second half of 2025; Cadastre (Parcels) theme, state grade tables, sub-grades, summary statistics and individual state report cards. https://nsgic.org/wp-content/uploads/2026/08/2025-GMA-Full-Report-20260814.pdf
2. National States Geographic Information Council, Geospatial Maturity Assessment initiative page, listing the 2025 Full Report and the assessment history, read August 2026. https://nsgic.org/initiatives/geospatial-maturity-assessment/
3. National States Geographic Information Council, 2025 GMA State Summary Table, 2026. https://nsgic.org/wp-content/uploads/2026/08/2025GMAStateSummaryTable.pdf
4. National States Geographic Information Council, 2023 Geospatial Maturity Assessment report, December 2023; Cadastre theme summary and program tables for the 35 states with parcel aggregation programs and the 9 without. https://nsgic.org/wp-content/uploads/2024/02/2023GMAReportFinal.pdf
5. U.S. Census Bureau, 2022 Census of Governments, Organization component, Table 2, Local Governments by Type and State, released 2023. https://www.census.gov/data/tables/2022/econ/gus/2022-governments.html
6. U.S. Census Bureau, 2022 Census of Governments data file CG2200ORG02, local government counts by type and state, 2023. https://www2.census.gov/programs-surveys/gus/tables/2022/cog2022_cg2200org02.zip
7. Federal Geographic Data Committee, Cadastral Subcommittee, charter description, public domain state standardization, standardized Public Land Survey System status and the Land Buy-Back Program for Tribal Nations parcel geodatabases, page read August 2026. https://www.fgdc.gov/organization/working-groups-subcommittees/cadastral/index_html
8. Federal Geographic Data Committee, Geospatial Data Act of 2018, signed October 5, 2018 as a component of the FAA Reauthorization Act, H.R. 302, Public Law 115-254, with the section cross-reference to 43 U.S.C. Chapter 46, page read August 2026. https://www.fgdc.gov/gda
9. Office of Management and Budget, Circular A-16, Coordination of Geographic Information and Related Spatial Data Activities, revised August 19, 2002, with the National Geospatial Data Asset theme history including the August 19, 2011 revision from 34 NSDI themes to 16 NGDA themes and the July 3, 2024 theme list update, via the Federal Geographic Data Committee. https://www.fgdc.gov/policyandplanning/a-16
10. 43 U.S.C. Chapter 46, Geospatial Data, sections 2801 to 2811, Cornell Legal Information Institute, current 2026. https://www.law.cornell.edu/uscode/text/43/chapter-46
11. Public Law 115-254, FAA Reauthorization Act of 2018, containing the Geospatial Data Act of 2018, enacted October 5, 2018, U.S. Government Publishing Office, printed 2018. https://www.govinfo.gov/content/pkg/PLAW-115publ254/pdf/PLAW-115publ254.pdf
12. Utah Geospatial Resource Center, Utah Parcels, SGID cadastre category, including per county last update dates for basic parcels and Land Information Records parcels and the LIR attribute list, page read August 24, 2026. https://gis.utah.gov/products/sgid/cadastre/parcels/
13. New York State GIS Program Office, Statewide Parcel Data, program created 2014, standardized tax parcel data for 38 counties updated annually, statewide parcel centroids and the Standardized Tax Parcel Data Dictionary, page read August 24, 2026. https://gis.ny.gov/parcels
14. Wisconsin State Cartographer's Office and Wisconsin Land Information Program, Statewide Parcel Map Initiative, V12 with 3.56 million records collected from counties in the first half of 2026 and V13 tentatively scheduled for June 30, 2027, page read August 24, 2026. https://www.sco.wisc.edu/parcels/data/
15. Montana State Library, Montana Spatial Data Infrastructure cadastral framework, monthly county cadastral downloads, Department of Revenue Orion computer assisted mass appraisal data and the cadastral REST endpoint, page read August 24, 2026. https://msl.mt.gov/geoinfo/msdi/cadastral/
16. Texas Geographic Information Office, Texas Water Development Board, StratMap Land Parcels, county availability, annual refresh statement and fitness for use statement, page read August 24, 2026. https://geographic.texas.gov/stratmap/land-parcels.html
17. Montana Code Annotated, Title 15, Chapter 7, Part 3, section 15-7-308, Disclosure of information restricted, exceptions, 2025 edition. https://archive.legmt.gov/bills/mca/title_0150/chapter_0070/part_0030/section_0080/0150-0070-0030-0080.html
18. Kansas Statutes Annotated, section 79-1437f, disposition and use of the contents of the real estate sales validation questionnaire, Kansas Office of Revisor of Statutes, current 2026. https://www.ksrevisor.gov/statutes/chapters/ch79/079_014_0037f.html
19. MMCG Research, U.S. Parcel Records Series: jurisdiction counts, digital and public access sums, locked jurisdiction counts, program characteristic tallies across the 2023 and 2025 state report cards, and the Utah per county vintage distribution, computed from sources 1, 4, 5, 6 and 12; MMCG database, 2026. https://mmcganalytics.com/methodology/