How much of a neighborhood’s spending stays in the neighborhood?

Look up any US place and see what residents spend on everyday things, and how much of it local stores capture.

Where do you want to look?

Try , , or . Or .

Who it’s for

Each opens South Shore, Chicago, as an example; search above for your own place.

The score, in dots

Every category gets a retention rate. A rate of 1.0 means local stores sell what the national pattern predicts for the people who live there. Lower means more of that spending goes somewhere else.

Local stores sell about 4 dollars for every 10 the national pattern predicts from residents’ spending. That is a retention rate of 0.4.

Illustration only.

Everything here is built from official statistics and other published sources: the American Community Survey, the BLS Consumer Expenditure Survey, the 2022 Economic Census, Nonemployer Statistics, LEHD LODES and ZIP Business Patterns, with store listings from USDA and Overture Maps and program and tax data from federal, state and local agencies and published surveys of state tax rules. It is an estimate of retail gaps, not a record of where people actually shop. Public beta; 84,415 census tracts covered.

Pick any neighborhood or group of census tracts in the US. See what residents spend in sixteen categories, what local stores sell in fourteen of them, and how that compares with the national pattern. Built from official statistics and other published sources, all documented.

Loading the figures…

Give this selection your own name, such as “71st Street corridor”. The name appears in the heading, the printed report, the citation and the names of files you download for it. A link to a place or tracts you picked carries the name, so anyone you send it to sees it; a suggested area’s link keeps the area’s own name. It changes nothing for anyone else.Name for this selection

Click a census tract to add it or take it out. Go to countyShade tracts by program

Loading…

No tracts selected

How to read this page

What it measures

For any group of census tracts, we estimate what residents spend in each category, using national spending patterns for households at their incomes, and what stores located there sell. Both come from official statistics: Census and BLS surveys of households, and Census counts of stores and jobs. It is not card data, so it shows where spending likely goes, not where residents actually shop.

How to read it

The retention rate is local store sales divided by what the national pattern predicts from residents' spending. 1.0 is the national norm, an average over every place in the country. Downtowns, malls and corridors that draw shoppers from elsewhere pull it up, so residential neighborhoods usually fall below it, while regional towns and small cities that serve the countryside around them often sit above it. The page therefore compares each area first with similar places (the same kind of area, setting and size across the US). A category in the bottom quarter of similar places is where residents' spending goes elsewhere; in the top quarter, the area draws shoppers in. More leaves than in the typical one is the dollars of residents' spending the typical similar place keeps but this area doesn't; leakage at the national norm is the same measure against the national average, shown as a second yardstick. Figures are ranges because store sales are estimated two ways: a range shows how far the two estimates differ, not a margin of error, and real uncertainty is wider. The reliability line under the summary says how far to trust an area's figures. Store counts say whether the gap is room for more stores or stores already here that are short of customers. Treat all of it as a planning estimate, not a sales forecast. Full method and limits

Biggest gaps first

The categories that rank in the bottom quarter of similar places, furthest short in dollars first. Each row of dots shows how many of every 10 dollars expected at the national norm local stores capture, to the nearest half, at the middle of the two estimates of local sales: filled dots stay here, ringed dots are the further dollars the typical similar place keeps, and hollow dots leave even there.

All categories at a glance

The retention rate: local store sales divided by what the national pattern predicts from residents' spending. 1.0 is the national norm, an average over every place in the country, which downtowns, malls and corridors that draw shoppers from elsewhere pull up: residential neighborhoods usually fall below it, while regional towns often sit above it. The small tick on each bar is the typical rate for similar places, and the color says where this area ranks among them. The solid bar is the lower of the two estimates of local sales, the lighter part runs to the higher one.

See all categories in a table

Annual, in 2024 dollars. "More leaves than in the typical one" is resident spending × (typical retention for similar places − retention here): spending the typical similar place keeps but this area doesn't; negative means this area keeps more. In the total row, the typical rate is each category's typical rate weighted by its expected sales, and the dollars add up each category's own gap: neither is the total of any one place. Leakage at the national norm is resident spending × (1 − retention rate); negative means the area draws in more than the norm. Each range runs between the two estimates of local sales: one placed by jobs at stores (LEHD), one by store payrolls by ZIP code (ZIP Business Patterns).

Tobacco and vaping is counted in the total but left out of the summaries, gaps and store lists. Child care is shown but kept out of the total: its sales include public payments (child care subsidies, Head Start) that concentrate in lower-income areas, while the spending survey counts only what households pay themselves, so its retention runs high there.

Spending public data can't match to local sales

Most medicine spending is paid through insurance at the pharmacy, and admissions are dominated by visitors. Vehicle purchases are left out entirely, as a category no neighborhood captures.

Stores: room for more, or spending elsewhere

Each bar: stores of the kind here for each one residents' spending carries at the rate similar places keep. Below 1.0, the area has fewer than its spending carries; that is room for more only when the category also ranks low among similar places.

How each reading is decided

"Carried at the national norm" is how many stores of the kind residents' spending in that category would carry at the US ratio of stores to spending. That norm is an average over every place in the country, so it sits above what most residential neighborhoods keep and below what many regional towns keep; "carried at the rate similar places keep" scales it by the median retention of places of the same kind, setting and size, and "room for more" is that figure minus the stores here. Sales per store compares sales per store here with the US average: well under 1.00 means the stores already here are short of customers, so residents' spending goes elsewhere even though the stores exist. Each ratio uses the same listing on both sides, so a listing that runs high everywhere cancels out. Map listings sort restaurants by cuisine, not service, so each tract's restaurants are split into full and quick service by the mix of restaurants in its ZIP codes (ZIP Business Patterns 2023), and the US counts the same way. Coffee shops, bakeries and dessert shops (cafés, juice and bubble tea, ice cream, candy) are a separate row; convenience stores and bodegas are counted with the food stores that take SNAP, under Stores here. These are planning estimates from public data, not a market study.

Compared with similar places

Similar places are areas of the same kind (city or town, county, or neighborhood), the same setting and a similar population, measured the same way across the US. Typical is their median retention rate, at the middle of the two estimates; the middle half runs from the 25th to the 75th percentile. Retention varies widely even among similar places, so read this as a reference, not a standard.

Compared with nearby areas, and side by side

Side by side with up to three other areas

Every area is measured the same way as the selection above and compared with places of its own kind, setting and size, so a large neighborhood and a small one are each judged against their own kind. "Here ÷ typical" is the area's retention over the typical retention of its similar places (each category's typical rate weighted by its expected sales); 1.0 means it keeps as much as they do. Similar places are grouped by kind, setting and population band, so the typical value can step at a band's edge, and two areas just either side of one are judged against different groups. Click an area to open it; the view and any areas side by side stay with it.

Find comparable places

Neighborhoods, towns or ZIP codes anywhere in the country that are most like this selection, in the ways you choose. Open any of them, or set up to three side by side with this one.

KindWhereSize Compare incomes in local prices

Each area is profiled from the same census-tract figures as the selection. Residents: median household income (by default in local prices: each median over its area's price level, from the Bureau of Economic Analysis's 2024 Regional Price Parities for its metro area or its state's areas outside metros (Delaware's statewide level for Sussex County, where the Bureau publishes none outside metros), so the same income buys about the same everywhere; untick the box to compare plain dollars), the shares of households under $35,000 and at $150,000 or more, people per household (counting those living in households, so dorms, barracks and prisons don't skew it) and median age (American Community Survey 2020–2024). Spending pattern: retention in each store category in the total other than tobacco, so areas whose residents' spending leaks from and holds in the same categories rank close; each strip shows that pattern. The dot plot places the selection and the results shown by percentile among the country's areas of the kind on each measure. Size: residents. Setting: a big city (a city of 250,000 or more residents, or an area at least half of whose households live in one), a metro area, a small-city region or rural, from the 2023 metropolitan and micropolitan delineation of the county holding most of its households. Each difference is measured against how widely that figure varies across the country's areas, so "very close" is within a quarter of that spread and "close" within half; the list ranks by the criteria that are on, equally weighted. Areas with fewer than 300 households are left out, and so are areas that share most of their households with the selection (the same place under another boundary, a neighborhood inside it, the ZIP code that holds it). "Compare the top 3" puts them in place of any areas already side by side. Comparable places are a starting point for finding peers to learn from, not a finding that the two will behave alike.

Where the spending goes nearby

Projected leakage of the area within 10, 25 and 50 miles of the selection (straight line from the nearest selected tract, the selection itself left out), at the middle of the two estimates; a surplus means its stores sell more than its residents' spending supports. "Balanced by" is the nearest ring whose leakage (or surplus) is at least as large as this area's surplus (or leakage). It shows where offsetting spending could come from or go, not where any one shopper goes, and it leaves out online spending and travel. Tracts that live off visitors (flagged as visitor economies on their own, with a large share of lodging jobs or of seasonal homes; see the method) add no surplus. "Main work areas" adds up the same figure over the tracts where residents most often work (LEHD LODES origin-destination jobs; see the method); it shows whether stores where residents work could absorb what leaks, not how much residents spend there.

More about this area

Open any of these; all of it is also in the report and downloads.

Sales tax on spending that leavesWhat it would pay in state and local sales tax, against similar places and the national norm

Each category's rate is the combined state and local sales-tax rate for the kinds of things it covers, averaged over the selected tracts by residents' spending: groceries are exempt or taxed at a lower rate in many states, gas is usually taxed by excise instead, and salon services and repair labor are taxed in some states and not others. Projected leakage is in survey dollars, which run below actual spending, so the tax figures lean low. Restaurant meals taxes are included where the state, county or city gives the rate, on drinks served too unless a city's code leaves them out; taxes by the unit or on the wholesale price (most excise taxes) and caps on single purchases are not (see the method notes).

Notes on this state's rates

Stores hereStores, restaurants, gyms, child care, clinics and hospitals counted here

Stores counted inside the selected tracts, against how common each kind is across the US per resident. Food stores come from USDA's list of every store that takes SNAP. The rest come from Overture Maps, an open listing of places that runs high for some kinds (salons, bars and home furnishings especially), so read those as indications, not a census.

Programs and designationsOpportunity Zones, New Markets Tax Credits, CDFI areas, HUD housing-credit areas, HUBZones, food access and state and local programs

How many of the selected tracts qualify for programs that bring investment, credit or food access, from the agencies' own lists. Eligibility is decided tract by tract (ZIP area or county for HUD's Difficult Development Areas), so a project's own address is what counts; confirm on the agency's map before relying on it. Shade any of these on the map from the menu above it.

Public and land-bank property

From the parcels large cities publish as held by a land bank, owned by the city or another public agency (surplus land included), or vacant. What each city publishes differs: some list only city-owned land, some a land-bank inventory, some vacant lots or buildings, so a zero can mean a city does not publish that kind. Owner names are not kept. Confirm a parcel's status with the city or land bank.

Near transit

Stations of subway, light rail, streetcar, commuter rail and monorail lines, and bus rapid transit stations where the agency reports the line as bus rapid transit (some rapid bus lines are reported as ordinary bus and are not counted), from the USDOT National Transit Map and the FTA National Transit Database. Measured here; not an agency designation.

State and local programs

Shares are the part of this area's land inside each district, measured on census tract shapes, so they are close but not exact at a district's edge. Designated sites, such as Pennsylvania's Keystone Opportunity Zone parcels, are listed rather than measured. Eligibility goes by a property's own address; confirm with the program before relying on it. The map can show each program from the menu above it.

Jobs located hereJobs at workplaces here by industry, and net commuting

Jobs at workplaces in the selected tracts, by industry, wherever the workers live (LEHD LODES 2024; Alaska 2016, Michigan 2021). Self-employed people are not counted. Rows marked store sales are the industries the jobs-based sales estimate starts from; it uses only the store types within them.

Then and nowResidents, households, income, jobs and food stores in 2008–2012, 2013–2017 and now

Take it with you

A report for the people you're sharing it with, the figures as spreadsheets, a link that reopens this area, or the figures as text for your AI assistant. The link keeps the report's choices too.

A report

Data and links

Copy for your AI copies this area's figures, notes, method and sources as plain text, with an instruction to answer only from them, for pasting into ChatGPT, Claude, Copilot or another assistant. The page sends nothing anywhere; what you paste is up to you, and the assistant's answers are its own.

Suggestions and corrections

An idea for the site, a figure that looks wrong for a place you know, or something that isn't working: send it here. Every message is read.

Method and limits

What this measures

This is a retail-gap estimate, not observed leakage. Local sales include shoppers who live elsewhere, and resident demand is modeled from national spending patterns by income. Neither side records where residents actually shop. That takes card data split by cardholder home and store location, which is the next layer (see the bottom of this page).

Census tracts

Census tracts are small areas the Census Bureau draws to hold about 4,000 people each (1,200 to 8,000). Every figure here is added up from tracts, so a named neighborhood, city or ZIP code is approximated by the tracts that lie mostly inside it; the fit note under each area says how closely.

Household income

Total household income is the ACS aggregate household income (B19025), before taxes, in 2024 dollars; where the Census does not publish a tract's total, it is estimated from the households in each income bracket, and the page says so. The median is estimated from the households in each income bracket (B19001), within the bracket that holds the middle household ("$200,000 or more" when it falls in the top bracket); the Census' own tract medians cannot be added up across tracts. The income chart compares the share of households in each bracket with the US. The Census Bureau withheld its 2020–2024 figures for 14 tracts in Suffolk County, New York (Brentwood, Wyandanch and parts of the towns of Islip and Babylon) after a data collection error in 2020 (ACS errata note 148); their figures here come from the 2018–2022 survey, the latest that publishes them, with incomes restated to 2024 dollars, and every page that includes one of them says so.

Resident demand

Residents are each tract's total population and households its occupied housing units (ACS 2020–2024 five-year, tables B01003 and B19001). Residents include people living in group quarters, such as dorms, prisons and nursing homes, who are not households and are not counted in demand; a tract that is mostly group quarters can show many residents and little or no demand. Demand is households by income bracket in each tract (B19001) times what US households at each income spend by category (BLS Consumer Expenditure Surveys 2024 and 2025, averaged, 2025 restated to 2024 dollars with CPI-U; auto repair from 2025 alone, since BLS moved vehicle parts, accessories and fluid to the Diary Survey in 2025, which measures them much higher), interpolated at each bracket's midpoint. The top bracket's mean is solved from aggregate household income (B19025), held between $200K and $1M. The total household income in the sentence at the top is that same aggregate (2024 dollars, before taxes; for the few tracts where the Census does not publish it, it is estimated from the income brackets): what residents earn, not what they spend, given only for scale, along with the share of it the measured categories take. Spending here is the model's estimate from the survey, which runs below actual spending, so the share leans low; where many households report little income (students, people living on savings, aid or benefits not counted as income) it can approach or pass all of reported income, as the survey itself shows at low incomes. Gasoline uses the 2025 survey's “Gasoline” by income (incomes restated to 2024 dollars with CPI-U), restated at 2024 pump prices. Gas, auto repair and vehicle purchases are scaled by the tract's vehicles per household (B25046) against what the survey shows at the same incomes, between 0.2 and 2.5 times.

Rural households spend differently from what their incomes alone predict. The survey's urban and rural tables (2015–2019, from the BLS series that ends in 2020; BLS's newer type-of-area series, from 2021, is not used yet) show them spending less on restaurants, alcohol, personal care and apparel, and more on tobacco, after income is accounted for two ways (the tracts' own incomes, and the survey's rural and urban mean incomes). Each tract's demand in those categories is multiplied by 1 − r + r × f, where r is its rural share (the part of its 2020 population outside Census urban areas) and f is the category's factor: restaurants 0.82, alcohol 0.60, personal care 0.74, apparel 0.78 and tobacco 1.78. A fully rural tract takes the whole factor; a fully urban one is unchanged. Groceries are not adjusted: rural households spend on food at home about what their incomes predict, and the survey leaves out food people grow themselves. A factor is used only where both ways put it at least 0.1 away from 1 on the same side and rural households spend less (or more) than urban ones in every one of the five years; the other categories are left unadjusted.

Local sales: two estimates

Jobs-based (LEHD LODES 2024): retail, food-service and personal-service jobs on each block, split into store types by the ZIP's industry mix.
ZIP-based (ZIP Business Patterns 2023): each ZIP's employment by store type, with each Census-suppressed size class given the fewest stores the suppression rule allows, each store counted at its size class's US average size for that kind of store (County Business Patterns 2023), and the remainder scaled to the ZIP's published sector total, then allocated to tracts by their share of the ZIP's jobs in that sector. ZIP Business Patterns leaves a kind of business unnamed where a ZIP has only one or two of them, and both estimates split a ZIP's jobs by the kinds it names, so unnamed stores are counted as the named kinds. Where some of a ZIP's businesses in a sector are unnamed (outside the named kinds, or inside a named kind beyond its named sub-kinds) and none of the named ones is a kind this site counts, the unnamed businesses are split, on both estimates, by the kinds of business Census leaves unnamed in the state's counties of the same kind, metropolitan, small-city or rural (County Business Patterns 2023 counts each county's businesses by kind wherever a county has three or more of a kind, and the state's count supplies the rest; what each county's ZIP codes name is taken away), each at its kind's typical size there (the middle of its size classes, or, for a kind too rare there to have one, such as department stores in rural counties, the middle over all counties), within the sector jobs the named kinds leave. A ZIP with only one or two businesses in a sector has no row for it at all, and some ZIP codes with jobs aren't listed at all (campuses, public or tribal employers, or businesses counted under a neighboring ZIP code); the ZIP-based estimate counts none there. Where any of this covers a quarter or more of an area's retail, service or restaurant jobs, the page says so.
Both held to each county's store jobs: neither source counts store jobs directly: workplace job counts put warehouses, headquarters, call centers and casino hotels in the same sectors as stores (a catalog seller's offices and warehouse made a small Wisconsin town look like a $1.2 billion shopping district), and ZIP Business Patterns gives only size classes. So both estimates are scaled, store type by store type, to County Business Patterns 2023 employment in each county (raised toward the Economic Census' US employment for that type by at most 5%: the 2022 Census codes count online and mail-order sellers in the store industries for what they sell, so its full counts would put online sales into stores), and they differ only in how they place a county's store jobs among its tracts. Where an estimate places fewer of a county's jobs of a type than County Business Patterns counts, what it placed is kept and the rest go to the county's tracts by the jobs that estimate placed there in the counted store types of that sector (retail, food service or personal service). What is spread is bounded by the 2022 Economic Census: a county's food and beverage stores, spread included, may sell at most 1.15 times the Census' county receipts for them (in 2024 dollars), in all and for each kind the Census publishes there; in Minnesota, where the Census withholds a county's liquor stores, theirs may reach the county's municipal liquor stores' 2024 sales (State Auditor) plus $1.96 million for each other liquor store, and no more. Jobs an estimate places itself are never cut. A county with only one or two businesses of a type has no row for it, and takes its share of the state's remaining jobs of that type, from a quarter of its own estimate up to its own estimate.
Fitness and child care: gyms and studios, golf courses, bowling centers, ski areas and other recreation, sports, dance and music instruction, and day care centers. Workplace job counts for these industries are mostly public schools, hospitals, casinos and museums, so both estimates use each ZIP's own ZIP Business Patterns rows for them (scaled so their US employment matches the 2022 Economic Census), placed in the ZIP's tracts by their share of its gyms and studios or child care centers listed on Overture Maps (or of its residents where it lists none); each ZIP's marinas are taken out of recreation. Their demand comes from the survey's detailed items by income: club memberships, fees for participant sports and recreational lessons; and babysitting, child care, day care and preschool. Self-employed child care (mostly home day care) and recreation are added by county, placed the same way.
Both estimates add self-employed businesses (Census Nonemployer Statistics 2024) for restaurants, bars, salons, barbers and other personal care, auto repair, coin laundries and dry cleaners: each county's receipts, placed in tracts by their share of the county's jobs in that industry. Self-employed retail sellers are left out, because most of their sales are online or direct, and so are food trucks, which the Census reports together with caterers.

Each store type's jobs are multiplied by its 2022 Economic Census sales per employee, taken from the place, county, metro area, state or US, whichever is the smallest with enough stores (pharmacies always at the US rate). A local rate is held to at most 1.5 times the state's rate, and a state's to at most 1.5 times the US rate: the 2022 Census counts online sellers within store industries and tabulates a firm's sales where it reports them, so places with a large online seller or headquarters show rates far above what their stores take in (general merchandise in the Seattle area 3 times the US rate, clothing in Nike's county 7.5 times). Sales are then split into categories as above, and restated in 2024 dollars (gasoline at 2024 pump prices). Thirty-four store types are used, from supermarkets and convenience stores to salons, dry cleaners, food trucks, gyms and day care centers; building-material stores, car dealers and online sellers are left out. The two estimates set every range on this page, except fitness and child care, where both use the same ZIP employment and give one figure. Alaska uses LODES 2016 and Michigan 2021, the latest each state has published.

Retention rate

Across the US, store sales run 0.9 to 3.1 times modeled resident spending, depending on the category. Most run above it, because households under-report spending in the survey and stores also sell to businesses, commuters and visitors. The categories that run closest to it are pets, toys and books (0.9) and child care (1.1); for pets, toys and books, much of the spending goes online or to store types left out here. The retention rate divides local sales by resident demand times that national ratio, separately for each estimate, so 1.0 means an area sells what the national pattern predicts for its residents. Projected leakage is resident demand times (1 − retention rate): residents' spending, in survey dollars, that local stores would keep at the national norm but do not. Because the survey under-reports some spending, it leans conservative. Store counts instead use the gap in store-sales terms (the predicted sales minus local sales), divided by a benchmark store's sales in that category.

Check against the South Shore baseline

For Chicago's South Shore (the City's 15 tracts), this page finds 25,043 households (CMAP's figure is 26,470) and $316M of resident demand in the five categories the hand-built South Shore baseline measured: everyday goods (groceries, household supplies, and pets, toys and books), restaurants, apparel, gasoline, and auto parts and repair. The baseline has $352M, $36M more: about $18M of that is the household count, and the rest comes from the baseline's survey item for gasoline (gasoline, other fuels and motor oil, where this page uses gasoline alone), its assumption for top-bracket income, and its single car-ownership adjustment to gasoline and auto repair for the whole area, where this page adjusts tract by tract. The baseline compared local sales with demand directly, found local sales at 63%–75% of demand and a gap of $88M–$129M; it could not measure home furnishings, personal care or alcohol from public data. This page measures 13 categories in its total against the national norm and splits store sales by what each kind of store sells, so its figures are larger and not directly comparable: a net projected leakage of $245M–$284M, with retention rates of 0.37–0.46 for groceries, 0.06–0.11 for full-service dining, 0.23–0.45 for quick-service food and 0.19–0.29 for home furnishings. The two agree on where the biggest gaps are: restaurants and everyday goods (here groceries, household supplies, and pets, toys and books).

Check against states' sales-tax records

Many states publish taxable sales by city or county and by kind of business. Where they do, the estimates here are compared with them.

Restaurants and bars are the cleanest comparison, since restaurant meals are taxed in every state with a sales tax. Store goods compare less directly: taxable sales leave out groceries in most states and include goods outside the categories here (electronics, sporting goods, jewelry and more), so the estimates usually come to less than taxable sales there. The test is skill against a naive baseline that predicts each area's sales from its number of households alone: areas differ a hundredfold in size, so a correlation of sales across areas is mostly a matter of size, and household counts alone already score high on it. Around each state's typical ratio of estimate to reported sales, the estimates' spread is a median 0.37 of the baseline's for restaurants and bars (0.18–0.83 across 30 states; below 1 is better) and 0.58 for store goods (0.30–1.76 across 25 states), and their correlation with sales per household, which household counts can't predict, runs 0.58–0.96 for restaurants and bars and 0.31–0.96 for store goods. For store goods in Arkansas, Colorado, Illinois, Nebraska, Nevada, North Carolina, South Dakota and Utah, the estimates correlate with taxable sales no better than household counts do, and in Colorado and Nevada their spread is wider than the baseline's too. The check is at city, town and county level: it does not test how sales are placed among a city's tracts, which neighborhood figures depend on, and a county check is partly met in advance, because store jobs are held to County Business Patterns' count in each county. The largest misses are small towns where tract lines cut through the commercial district, warehouse towns, and places whose sales come mostly from hotels and conventions, like Rosemont, Illinois.

State by state: sales-tax rates and checks

Data and dates

Each main source covers a set period and is replaced when its agency publishes a newer one. The table gives the period each covers, the day it was last checked for a newer release, and when the next is expected. Where that date has passed, the row says a newer release may be out and not yet added.

When each main source was last updated and checked

Similar places, the surrounding region and visitors

Similar places: every Census place (the tracts with at least half their residents in it), county and named neighborhood in the US with at least 300 households is grouped by kind, setting and population; neighborhoods are filtered as the search filters them, and neighborhoods that cover the same tracts count once. Setting is the county's: in a metropolitan or micropolitan area, or neither (OMB delineation, July 2023); a neighborhood's setting is whether a city of 250,000 people or more holds at least half of its households (the same rule as Find comparable places). Groups with fewer than 30 areas are merged with a neighboring population band. Small towns count only when they hold at least half of a tract's residents, so the groups for towns under 2,500 people are small (about 200–650 places) and lean toward towns that fill a whole tract. A ZIP code or hand-picked selection is compared as a neighborhood when at least half of its households live in such a city, or in a smaller city with at least twice its residents (then with neighborhoods of smaller cities), otherwise as a city or town; a suggested area is compared as a neighborhood. The page leads with this comparison: a category's gap against similar places is resident spending × (their median retention − retention here), each counted as at most 1.0: a rate above 1.0 means a place also draws in shoppers from elsewhere, which is not its residents' spending, so the gap compares only how much of residents' own spending stays and never exceeds what they spend; and its rank among them (bottom quarter, middle half, top quarter) decides whether spending reads as going elsewhere or drawing shoppers in. The national norm stays beside it as a second yardstick: higher than the typical one for most neighborhoods, lower for many regional towns and small cities. A total against similar places adds up each category's own gap; it is not the total of any one typical place.
Find comparable places ranks every named neighborhood, Census place, town or township and ZIP code in the US with at least 300 households (areas covering the same tracts count once; those sharing most of their households with the selection are left out) by how far it is from the selection on the criteria the reader turns on, each measured from the same tract figures as the selection. Residents: median household income (by default in local prices, over the Bureau of Economic Analysis's 2024 Regional Price Parity, all items, of the area's county's Metropolitan Statistical Area, or of its state's nonmetropolitan portion, which includes micropolitan counties (BEA prints none for Delaware's, so Sussex County, the state's only county outside a metro, takes Delaware's statewide parity); the shares below stay in the survey's dollar brackets) and people per household (both on a log scale; people per household counts residents living in households, ACS B25008, leaving out group quarters such as dorms, barracks and prisons), the shares of households under $35,000 and at $150,000 or more, and median age (each tract's ACS B01002 median, averaged over the area weighted by residents). Spending pattern: retention in each store category in the total except tobacco, on a log scale floored at 0.05. Size: residents, on a log scale. Each difference is divided by how widely that figure varies across all the areas listed (its standard deviation), and a criterion's distance is the root mean square of its differences. Setting counts each step between a big city (a city of 250,000 people or more, or an area at least half of whose households live in one), a metro area, a small-city region and a rural county as one; a Census place that is smaller takes its county's setting. "Closest on" names the closest of the criteria measured against a spread (not setting) when two or more are on. The ranking is the root mean square of the criteria that are on. Incomes in local prices are measured against the spread of incomes in local prices. Words: "very close" is within a quarter of the spread, "close" within half, "somewhat alike" within one.
The surrounding region sums the projected leakage of every tract whose interior point lies within 10, 25 or 50 miles of a selected tract's, including neighboring states. A tract adds no surplus when it flags as a visitor economy on its own (below) and lives off visitors: its hotel and other lodging jobs are at least 10% of the jobs located there (about 8.3 times the US share) or its seasonal homes at least 3 times the US share. Lodging jobs by tract are an estimate: each county's lodging employment (Census County Business Patterns) spread to its ZIP codes by their lodging businesses (ZIP Code Business Patterns) and to tracts by their hotel and restaurant jobs (LODES), among the ZIP's tracts that hold a lodging listing other than a campground, RV park or rental in Overture's places data, at Overture's confidence of 0.5 or more (the lodging places counted on the pages use 0.9), when any does. So a resort's or casino's sales to visitors aren't read as absorbing a neighborhood's leakage, while a downtown whose hotels sit among many offices, or a mall's restaurants, stay in (the visitor score alone would flag 23% of all tracts, mostly restaurant and shopping tracts that serve residents; with the visitor signal it flags 10%). ZIP Code Business Patterns leaves out an industry with fewer than 3 businesses in a ZIP code, so a resort alone in its ZIP has no row and its county's lodging jobs go to the county's other ZIP codes, and some resorts are missing from Overture's listing: a few casino and resort tracts show few or no lodging jobs and aren't flagged, while a few others show more than they hold (each tract is capped at its LODES hotel and restaurant jobs).
Where residents work comes from the LEHD LODES origin-destination files (2024; Alaska 2016, Michigan 2021; all jobs, every state's residents followed across state lines): each tract's jobs held by its residents, by straight-line distance from home to the work tract, and the tracts holding at least 1.5% of a home tract's jobs (up to eight), whose leakage and surplus are added up as "main work areas". Distances run between tracts' interior points, so in very large tracts (much of Alaska and the rural West) a job can show as farther from home than it is; and LODES places some jobs at an employer's reporting address rather than at the site where the work is done, so it shows less commuting within a county and longer commutes than the American Community Survey does for the same workers (Census Bureau Center for Economic Studies working paper 17-34, 2017): read the farther shares as high. A tract's residents who work in another state are counted in that state's LODES year. LODES counts jobs, so a resident with two jobs counts twice. Where the main work areas are compared, the page also gives an upper bound for restaurant spending near work: residents' restaurant spending × the share of restaurant spending that is weekday lunch at fast-food and full-service restaurants bought by households with a worker (16%: in the Consumer Expenditure Diary of 2021, the last year it recorded meal type before its 2022 redesign kept only the kind of place, lunch was 28% of working households' weekday fast-food and full-service spending; 2021 was a year of more working from home than usual, so the share may run low: the 2019 diary would give 17%; that part is assumed to hold, and is applied each year to the weekday share of working households in the current diaries (2024 and 2025); vending is left out throughout) × the share of residents' jobs outside the area. It also counts lunches by household members who don't work, on days off and near home, so it is a rough upper bound (rough because the 2021 share may run low, above); it is never taken out of the leakage figures. No such share exists for stores, so for stores the page says only which leaks the stores where residents work could absorb.
Border notes: a neighboring state is compared where its land lies within 20 miles of a selected tract's interior point and a county of it that close can be reached by land: it touches the tract's own county, or a county that touches the tract's county and is itself that close. Counties less than a mile apart count as touching, so a river with its bridges joins them; a wider crossing doesn't (the Potomac at the Nice Bridge, about 1.7 miles, drops out), and neither does open water (Long Island and Fishers Island aren't compared with Connecticut); a state two or more counties away isn't compared even when its land is that close (Queens and upper Manhattan with Connecticut). The distance given is a straight line and can cross a bay or a river, so the drive can be much longer (Bridgeton, New Jersey, to Delaware across Delaware Bay; Greenwich, Connecticut, to New Jersey across the Hudson). The comparison covers gas, tobacco and alcohol bought in stores. Gas: state taxes and fees a gallon of gasoline from the Energy Information Administration's table of motor fuel taxes as of July 1, 2026 (its total of state tax and other taxes and fees, which carries sales taxes on fuel where a state collects them by the gallon, New Jersey's petroleum gross receipts tax as a rate a gallon, and Florida's 6-cent local option fee, as EIA counts them), changed where a state's own page, or EIA's own note, shows a different current rate (Illinois, North Carolina, New York, OH, Utah and Vermont; New York's state sales tax on fuel, 8 cents a gallon and 8.75 in the Metropolitan Commuter Transportation District, which EIA lists outside its total, is added at 8 cents). County and local fuel taxes are left out, and so are the small taxes Delaware and Ohio levy on petroleum gross receipts, which EIA gives only as percentages; both states are still compared on the rest. Gas isn't compared with a neighbor whose stores within reach charge a local sales tax on fuel of half a point or more (California, Georgia, Illinois and New York, as their rate tables carry it), since that tax could close the gap; Connecticut isn't compared, since its 8.1% petroleum gross earnings tax isn't a rate a gallon. Cigarettes: each state's tax a pack, confirmed on its own page, plus the local taxes a pack that cities and counties levy where their governments or the state publish them (418 in Alabama, Alaska, California, Colorado, Illinois, Missouri, New York, Ohio, Pennsylvania and Virginia), each tract taking its city's and county's as the county's rules allow; local taxes not confirmed on an official page, or set as a share of the price, are named in the note but not counted, and Missouri's other city cigarette taxes (there's no statewide list) and Alabama's police-jurisdiction taxes outside city limits aren't counted. Sales tax on tobacco, and on alcohol bought in stores, comes from both states' rate tables (here, residents' own tracts weighted by their spending; there, its tracts within reach, weighted by their stores' sales in the category); on alcohol it includes local taxes charged on the price where the rate tables count them (Chicago's 1.5% on bottles), but not per-gallon taxes or the taxes states collect in their own liquor stores, so alcohol isn't compared with Pennsylvania (its 18% liquor tax on wine and spirits in state stores) or Virginia (20% on spirits sold by its ABC stores, Va. Code § 4.1-234) as the neighbor. A difference of about 3% of the price or more (3 points of sales tax, 10 cents a gallon, 50 cents a pack) is noted. No volumes are estimated: gas prices swing too much month to month.
Visitor economy: the measure USDA's Economic Research Service uses for its recreation counties (County Typology Codes, 2025 edition; after Johnson and Beale). A whole county is flagged when ERS classes it as high recreation. Any other area is scored the same way on the public figures below the county: jobs in arts, entertainment and recreation, hotels and restaurants, and real estate and rental as a share of all jobs (LODES 2024; Alaska 2016, Michigan 2021; weight 0.6, since earnings by industry aren't published below the county and ERS weights earnings and jobs 0.3 each) and homes held for seasonal or occasional use as a share of all homes (ACS 2020–2024; weight 0.4), each measured against the mean and spread of nonmetro counties on the same figures; the area is flagged at ERS's cutoff, a weighted score of 0.67 or more, when a sign of visitors joins the score, since below the county a high share of restaurant and real-estate jobs can come from a small job base in a bedroom suburb: lodging jobs (by tract an estimate, as described above) at least twice the US share (2.4% of all jobs), seasonal homes at least the nonmetro county average (8.3% of homes), or at least 5 lodging places and as many per resident as the top 5% of places with 1,000 or more residents (2.2 per 1,000 residents). Scored this way, counties agree with ERS's own codes 94% of the time, most of that among the many counties neither flags: the score finds 295 of ERS's 431 high-recreation counties and adds 47; LODES leaves out the self-employed, so the score runs a little under ERS's for places with many proprietors. Where an area isn't flagged but has at least 5 lodging places and as many per resident as the top 5% of places with 1,000 or more residents (2.2 per 1,000 residents), or at least 50 seasonal homes and at least 3 times the US share of homes (9.9%), the page shows a visitor caution instead of the flag (the site's earlier visitor test); a whole county ERS doesn't class gets the caution when the score and a visitor signal, measured the same way as below the county, would flag it.

What this site keeps

The first download in a browser asks for an email address; role, organization and an updates opt-in are optional. They go to the site's spreadsheet, with what was downloaded and the area; only the site's owner and the people they ask to review suggested areas can open it, and the first download from an address is also emailed to the owner. This browser remembers the address. Messages and suggested areas go to the same spreadsheet. The site also counts which areas are looked at and downloaded (the date, state, area name and number of tracts), with no address or identifier, though a counted download can be matched to its download record by date and area. Emails about new data, features and paid plans go only to addresses that have confirmed they want them. Page visits are counted by Cloudflare Web Analytics, which sets no cookies and keeps no personal data. Nothing is sold or shared.

Jobs, stores and programs

Jobs located here come from the same LEHD LODES files, by workplace and all 20 industry sectors. Jobs held by residents come from LODES residence files, wherever those jobs are. Both count jobs, not people, so their difference is net commuting. Self-employed people are not in either. LODES publishes workplace jobs by census block with noise added to protect employers' confidentiality, and an employer's jobs can be placed in a different block from one year to the next, so figures that rest on where a tract's jobs are can change noticeably between LODES years where a tract has few jobs or a large employer's jobs are placed in another block: each tract's jobs-based store sales, where a ZIP code's store sales are placed among its tracts, a tract's blended sales-tax rate (below), and so retention for small areas. The figures here follow the LODES files as published (2024; Alaska 2016, Michigan 2021).
Hospitals are every hospital on Medicare's list (CMS Hospital General Information, October 2026), placed by the Census Geocoder or, where it can't place the address, by matching the name to a hospital listing on Overture Maps in the same ZIP code (or the same county, where the ZIP is a post-office box or single-site ZIP with no area of its own); about 5% can't be placed either way and are left out. Map listings alone count a hospital's departments and offices as separate hospitals. Community health centers are HRSA's active, permanent health center sites, including school-based ones (not mobile vans or seasonal sites). Doctors' offices and clinics, dentists, urgent care and mental health clinics are Overture Maps listings, private and public alike.
Stores here: food stores are USDA's list of every store authorized to take SNAP (downloaded October 2026). Other kinds are places listed in Overture Maps (October 2026 release, places with a confidence score of at least 0.9 and not marked closed), an open dataset compiled from several map sources; it lists more salons, bars and home furnishings stores than Census business counts, so treat those rows as indications. National rates per 10,000 residents use the same sources. The store section splits each tract's listed restaurants into full and quick service by the mix of employer restaurants in its ZIP codes (ZIP Business Patterns 2023), since the listings sort restaurants by cuisine, and the US counts the same way; it compares gyms and studios by count only, since golf, bowling, skiing and lessons are not listed.
Sales tax on spending that leaves leaves out fitness and child care, which states tax very differently (gym memberships in some, lessons in fewer, child care almost nowhere) and which the rate tables here do not cover yet. It applies each state's combined state and local sales-tax rates, category by category, to the spending that leaves beyond what the typical similar place loses (only in the categories that fall short of it) and, beside that, to projected leakage at the national norm: the rates where the selected tracts are, and only on what each state taxes. Sales tax is charged where purchases are made, so a tract split between places with different rates (a city or town with rates of its own and the area around it, two county parts of a city, or two towns) takes each one's rates for the share of the tract's jobs located there: its store jobs for purchases in stores, and its jobs at restaurants, bars and hotels for restaurant meals and drinks served (LEHD LODES 2024; Alaska 2016, Michigan 2021 workplace jobs by census block); a tract with no such jobs takes them by its residents (2020 census), and the notes say where that happens. LODES gives jobs by sector only, so store jobs here are all retail trade jobs, car dealers, gas stations and the warehouses of online and catalog sellers among them, and the restaurant count includes hotel jobs; a tract's split can lean toward a car lot or a hotel that sells little of what a category covers. A city's or town's own page uses its own rates instead: sales inside a city are taxed at the city's rate, so the page takes the rates inside it, weighed across its county parts and towns by its own store jobs (restaurant, bar and hotel jobs for meals and drinks served; its residents where it has none), with a district's higher rate shown as a range as below, whatever the tracts around it charge. An unincorporated community's page (a Census designated place) takes the rates inside it the same way. The page says whose rates they are: the city's or town's own where it has a rate of its own on the state's list, or else those in force where it lies, naming the county (or the town with rates of its own), as for an unincorporated community, which levies none, or a city or town with no rate of its own (Florida's cities, New York's towns), or the state's alone where no local rate applies there; a city whose only levy is a restaurant meals tax is named for that. Any other selection (a neighborhood, a ZIP code, a county, tracts picked on the map) takes each tract's rates as described here, and the page says which applies. In Washington, where transit districts set different rates within a county, each tract takes the average rate over its stores (restaurants, bars and hotels for meals), in cities or not, from the Department of Revenue's location-code map. Groceries, gas, salon services and repair labor are exempt or taxed differently in many states; personal care, apparel and auto are split between goods and services by the services' national share of sales, and alcohol between bottles bought in stores and drinks served by their national shares of sales (2022 Economic Census product lines), each part taking its own rate. Auto repair shops' receipts take the treatment of repair labor: the Economic Census reports the parts they install inside the repair line, so where a state taxes installed parts but not labor, the tax on auto repair is a lower bound. Spending won back from stores elsewhere in the same state already pays the state's share there, so the page leads with the local share, the area's own gain, and calls the state's share mostly moved rather than new. Online orders are a different case: in states that tax a delivered order at the rate where it is delivered (destination sourcing, the rule of the Streamlined Sales Tax agreement), an order shipped to a home here already pays the local rate here, so the part of a gap that is online shopping beyond the national norm would bring in little new local tax if won back. Rates come from each state's revenue department (or the Streamlined Sales Tax files its members publish), as of October 2026. A city in more than one county takes, in each tract, the rates of its part in that tract's county. Where part of a place or county carries a higher rate, from a transit or special district, the rate shown runs up to it as a range. Restaurant meals taxes are included: state meals taxes in Connecticut, DC, Maine, New Hampshire and Vermont, local ones where a state publishes them (Massachusetts, Mississippi, Rhode Island and Utah) and, in Alabama, Arizona, Arkansas, Colorado, Florida, Idaho, Illinois, Indiana, Kentucky, Louisiana, Maryland, Minnesota, Missouri, Montana, Nebraska, New Jersey, North Carolina, North Dakota, Oregon, South Carolina, South Dakota, Tennessee, Vermont, Virginia and Wisconsin, where the state, county or city gives the rate; each state's notes name them, and they apply to drinks served as well as food, except Arkansas' local restaurant taxes, 11 Kentucky cities' restaurant taxes, 4 North Dakota cities' and Ramsey County's restaurant taxes, the restaurant taxes of Ashland, Cannon Beach and Yachats (Oregon) and the restaurant tax of Eagle Butte (South Dakota), which leave drinks out (the notes say so). Local meals taxes whose rate no state, county or city publishes are not counted. One rule decides what else counts: a tax charged as a percent of the retail price, whether the buyer pays it at the register or the seller pays it on its retail receipts, is counted; a tax by the unit (per pack, gallon or liter) or on the wholesale price, paid before the retail sale, is not, nor are lodging, car rental and admissions taxes or caps on single purchases. A percent-of-price tax that falls on only part of a category is counted on that part's share of sales where an official figure gives one: nationally, drinks served by kind (spirits 43%, wine 27% and beer 30% of drinks served), liquor stores' share of alcohol bought in stores, for a tax on what one kind of store sells (47%), and jewelry and watches' share of clothing, shoe and jewelry sales (13%), where clothing is exempt but jewelry isn't (Massachusetts, Minnesota, New Jersey, New York, Pennsylvania, Rhode Island and Vermont), all from the 2022 Economic Census product lines; or the state's own official figure where it publishes one: beer's share of liquor stores' alcohol sales in 2024, from Rhode Island's Division of Taxation. Otherwise it is named in the state's notes and not counted: taxes on vapor products (no official figure gives their share of tobacco sales), taxes on spirits or wine bought in stores where the state gives no split (the Census doesn't split alcohol bought in stores by kind, and liquor stores' share counts beer and wine shops too), and, where clothing is exempt, the part of an item's price over a cap and other taxed accessories. Counted this way, beyond each state's general and restaurant-meals rates, are state taxes on alcohol served in Arkansas, Kansas, Maryland, Minnesota, North Dakota, Oklahoma, South Carolina, Tennessee, Texas and Vermont, and bought in stores in Arkansas, the District of Columbia, Kansas, Maryland, Minnesota and North Dakota, and local taxes on alcohol or tobacco in Alabama (Jefferson County and 35 cities and towns); Alaska (Anchorage Municipality, Denali Borough, Fairbanks North Star Borough, Juneau City and Borough and 13 cities and towns); Arkansas (Fayetteville, Fort Smith, Little Rock and Sherwood); Colorado (Boulder); Georgia (24 cities and towns); Illinois (Chicago, Evanston and Rockford); Mississippi (Hattiesburg and Oxford); New Jersey (Atlantic City); Pennsylvania (Allegheny County and Philadelphia County); South Dakota (93 cities and towns); Wisconsin (9 cities and towns); cigarette and gallonage excises are not. Where a state's local rates differ by kind of purchase, as New York's do for clothing, New York City's for personal services and motor fuel, Georgia's for motor fuel, Arizona and Idaho cities' for groceries and restaurants, Denver's, Boulder's and Telluride's for food and drinks served to eat there, and Oxford, Alabama's for alcohol, each kind takes its own rate. Gaps for each state are listed with the figures and in the table below.
Programs and designations come from the agencies' tract lists: Treasury's 2018 Opportunity Zones and its 2027 zones, published by the CDFI Fund on November 5, 2026 and in effect from January 1, 2027 (6,576 tracts, 6,325 of them in the 50 states and DC, 2,439 of those rural); the CDFI Fund's New Markets Tax Credit and CDFI Investment Area eligibility (2016–2020 ACS); USDA's food access maps (2019 supermarket access; 2025 access to any SNAP store); and HUD's Qualified Census Tracts and Difficult Development Areas for the Low-Income Housing Tax Credit (2026 and 2027 lists). Difficult Development Areas are drawn by ZIP area in metro areas and by county elsewhere, so a tract counts when at least half its land lies in one (a ZIP area split between HUD metro areas counts only when every part of it is designated). Lists drawn on 2010 tracts are carried to a 2020 tract when at least half its land lies in listed 2010 tracts. "Severely distressed" here applies the three main NMTC tests only; a tract can also qualify in other ways. HUBZones come from SBA's HUBZone map service (its current Qualified Census Tracts and Non-Metropolitan Counties, governor-designated and disaster areas, and Indian lands, a tract counting for Indian lands when at least half of its area is inside).
State and local programs (enterprise zones and other state incentive areas, tax increment financing districts, special service areas and business improvement districts, brownfield areas, state designations such as Main Street, cultural and downtown programs, transit-oriented designations, and designated sites) come from the official boundary layers that states, counties, cities and regional agencies publish, read when the site is built. New Jersey's Qualified Incentive Tracts are computed here from the statute's test, since NJEDA's layer can't be republished; a layer whose terms bar or restrict republishing is left out, and the state's coverage line names it with a link to the publisher. Each district's share of a selection is the part of the selected tracts' land inside it, measured on the generalized 2024 tract shapes in an equal-area projection (shares that round to 1% of a tract or more); a program's total counts land in two overlapping districts once, and districts past their last year in force are left out. Designated sites (points) are listed by the tract they sit in. Coverage is growing place by place, and the programs section names the sources covered in the selected state, so a missing district elsewhere does not mean there is none. South Carolina, Mississippi, Arkansas, Tennessee, North Carolina and Georgia rank every county into tiers for job tax credits and similar incentives, Virginia's Commission on Local Government classes every county and independent city by fiscal stress, Washington lists its distressed counties (Employment Security Department) and its rural counties (Office of Financial Management), and Alabama's targeted and jumpstart counties are Enterprise Zones county-wide; in those states every county a selection touches is listed with its tier or designation from the state's current list, with its share of the selection's households.
Public and land-bank property counts the parcels that 49 large cities publish in their official GIS layers as held by a land bank or land reutilization authority, owned by the city, county or another public agency (surplus land included), or vacant (citywide vacant-lot or vacant-building layers). Where an owner field mixes public agencies and private owners, only public and land-bank parcels are counted, and no owner names are kept. A parcel listed twice counts once, in the tract that holds it. Each city's layer is taken as it stands, so what is counted differs by city and a tract outside these cities has no count; a city list from before 2024 is counted with its year shown, and the page names the layers each area's counts come from. Charlotte's layer is linked rather than counted, because its terms limit reuse.
Near transit is measured here, not taken from an agency list: the share of the selection's land within a half mile (804.67 m) of a station, using the USDOT National Transit Map (agencies' GTFS schedules) for subway, light rail, streetcar, commuter rail and monorail stations, and the FTA National Transit Database for bus rapid transit stations and for rail stations of agencies whose schedules the map lacks. Platforms and entrances of one station count once. Cable cars, aerial trams, funiculars and the Alaska Railroad are left out. Bus rapid transit is counted only where the agency reports the line as bus rapid transit, so some rapid bus lines reported as ordinary bus are missing. A half mile is the common station-area distance (FTA, 76 FR 52046).

Neighborhoods and other areas

Neighborhoods outside Chicago come from Who's On First, an open gazetteer. In some cities its boundaries come from the city's own planning or GIS office (Atlanta, Baltimore, Boston, Denver, Los Angeles, New Orleans, Oakland, Portland, San Diego, San Francisco, Seattle and others) or from published maps (Philadelphia, Washington, New York); elsewhere most come from Zetashapes and Quattroshapes, algorithmic boundaries assembled from crowd-sourced place data. Each result names its source. There is no official national neighborhood map, so coverage and quality are uneven. Boundaries are kept with a warning when the tracts trace them poorly, except that algorithmic ones whose tracts trace them under 20% (fit, as the page reports it) are left out in cities and towns of fewer than 250,000 residents (an area filed under a county instead of a city: in counties of fewer than 250,000), where the figures would mostly describe someone else's blocks; their names stay searchable as names only, with no tracts chosen, so a person who knows the area can pick its tracts (none where a city or town of the same name covers or borders the same tracts, or where a named area of the same name has a boundary in the same city; a few names that match a city or town in the same county but sit in tracts that don't touch or border it are kept, since the data can't tell whether they are the same place); other algorithmic ones can still be much smaller than a census tract, so their figures describe the whole tract around them, as the page says. Anyone who knows an area can pick its tracts on the map, rename it, or suggest it as a named area; the search always offers to build the name typed by hand. Where one city has two overlapping boundaries with the same name, the City's or a city office's own boundary is kept, then the one the tracts trace better; same-name places that do not overlap, such as two Koreatowns in New York, are both kept.
Chicago has the City's 77 community areas (from the City's 2020 tract assignments; eight tracts missing from that file, three of them in South Shore, are placed in the community area that holds most of their area) and the City's 98 neighborhoods; a few, such as Magnificent Mile, are much smaller than a tract and carry the same warning. Names in common use that are not official, such as Pilsen, come from Who's On First, and searching one that sits inside a community area, such as Pilsen in Lower West Side, also offers the community area.
Fit measures how well the tracts trace a boundary, not whether the boundary itself is right: their overlap as a share of the area and the tracts combined (100% is exact). Each area takes the tracts with at least half their area inside it, or the one tract that covers most of it. The page also says what share of the area the tracts cover and what share of the tracts lies inside it.
ZIP codes are Census ZCTAs: the tracts with at least half their land in the ZCTA, or the one tract that covers most of it, with fit measured on land area. They will not match postal ZIPs exactly.
Cities and towns are Census places: the tracts with at least half their residents in the place (2020 census block counts; census blocks for a tract with no residents). A town that holds less than half of the residents of every tract it touches (most towns under a few thousand people, and a few spread-out places) opens the tract that holds the most of its residents, and the page says the figures cover that whole tract and, when that tract holds under half the town, how much of it. Where a place's own tracts hold under half of its residents, or have no households (a prison, campus or base tract), the tracts holding the most of the rest of its residents are added until the page holds at least half of them. In the twelve states where towns and townships are local governments (Connecticut, Maine, Massachusetts, Michigan, Minnesota, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont and Wisconsin), a town or township that isn't also a Census place (Edison Township, New Jersey; Middletown, Rhode Island) is found the same way, from the residents of the census blocks in it.
Suggested areas: anyone can suggest a selection of tracts as a named area. It shows in their own search on that device at once, and in everyone's once it is reviewed and approved, labeled as a suggestion. The page does not ask who is suggesting and saves nothing about them.
Boundaries are 2024 Census cartographic files (2020 tracts), simplified for the map. Tracts that are only water have no households and are not drawn.
Streets on the map come from OpenStreetMap. Pointing at a tract (or tapping it on a phone), zoomed in to street level, lists the named streets that run along its edges, matched within about 45 meters of the simplified tract lines, so now and then a name is missed or a nearby parallel street is listed. Only streets in the part of the map on screen are checked.

Limits

Small areas are noisy. Jobs are reported by workplace block and can be miscoded, and a single large store can swing a category.
Self-employed retail sellers and food trucks are not counted (see Local sales above), which overstates leakage a little where those are common.
Product-line shares are national averages for each store type; a local store can sell a different mix.
Pet spending in the survey includes veterinary care, but vet clinics are not among the store types, so local sales of pets, toys and books lean low.
Where a ZIP's figures hide which kind of store holds its jobs (for example, supermarket or convenience store), the jobs are first split by the national mix of those store types; holding each county to County Business Patterns' count of each type then sets the county's mix, but not how it falls among the county's tracts.
Sales per employee and the benchmark store's sales come from the 2022 Economic Census store industries, which code online sellers by what they sell, so for some store types they pool online sellers with stores, raising some types' figures and lowering others'; the cap on local rates trims only the extremes.
Holding each county to County Business Patterns' store jobs raises an estimate by at most 5% for a type (the 2022 Census' higher counts include online sellers) but lowers it without a floor where County Business Patterns counts fewer, so a type whose industry code changed between the 2017 and 2022 classifications (convenience stores, food trucks) can lose jobs that moved to another type.
Online spending is inside every gap; no neighborhood captures all of it.
The ranges show only how far the two supply estimates disagree, and the two share inputs (each uses the ZIP's industry mix and tract job shares); they are not a margin of error. Demand is a single estimate. Real uncertainty is wider than the ranges.
Every selection carries a reliability label, from its households and census tracts. The survey behind residents' spending samples households, so spending modeled from it carries sampling error: for a typical single tract, a 90% margin of about ±13–21% on its spending in the store and restaurant categories, from the ACS margins of error on its household count (the lower figure) and on its households by income bracket, summed by the ACS Handbook's method, which runs high (the upper figure); 55% of tracts are above ±20% by the upper figure. That is about as large as the model's own regional error (below). The margin shrinks with the square root of the households (checked on the 1,310 counties of 5,000 to 20,000 households, summed tract by tract: a median of ±7% against ±8% from the scaling). Low: a single tract or fewer than 1,500 households (about ±13–20% at 1,500), where one store can also swing a category. High: at least 10,000 households in 5 or more tracts, where sampling error is about ±5–8% or less, below the model's regional error, which is then the larger part. Medium: everything between. The label states the selection's own sampling margin, scaled from the tracts' median. Error in the spending survey's own averages and the model's error come on top.
Residents' spending uses national spending curves by income in every tract; regional prices and habits are not adjusted for. BEA's state consumer spending (2024) shows how far states depart from those curves: between the 10th and 90th percentile of states, spending runs groceries 0.87–1.14, restaurants 0.82–1.11, apparel 0.83–1.10, home furnishings 0.87–1.09, gasoline 0.73–1.23 and personal care 0.66–1.18 times the national pattern, roughly ±11–26% by category. A state's departure passes into its retention rates and its places' ranks among similar places. BEA's state figures also count shopping across state lines, so they are not applied; spending curves by region and income, which BLS publishes, are the planned upgrade.
For a county or a larger selection the two supply estimates nearly agree, because both are held to County Business Patterns' store jobs in each county, so the range there shows almost none of the uncertainty.
Tobacco sales include cannabis where it is sold legally; the survey's tobacco line does not.
Totals net categories against each other; read the category rows before siting any one kind of store.
The Economic Census is from 2022 and is restated in 2024 dollars, not re-surveyed.

Sources

American Community Survey 2020–2024 (B19001, B19025, B25046, B01003)
BLS Consumer Expenditure Surveys 2024 and 2025, including the Diary public-use files; BLS Consumer Price Index (CPI-U); EIA retail gasoline prices
LEHD LODES 2024 workplace jobs
ZIP Business Patterns and County Business Patterns 2023
2022 Economic Census, including products by industry
Who's On First neighborhood boundaries (CC BY 4.0 and the licenses of their sources)
Census ZCTA–tract relationship file and 2024 cartographic boundaries
Census Nonemployer Statistics 2024 (self-employed businesses) and LODES residence files
USDA SNAP Retailer Locator, and places from Overture Maps Foundation (CDLA Permissive 2.0; includes Foursquare Open Source Places, Apache 2.0: attribution, notice)
Treasury Opportunity Zones and the CDFI Fund's list of designated zones, CDFI Fund eligibility (New Markets Tax Credits, CDFI Investment Areas) and the USDA Food Access Research Atlas
Taxable sales by city or county and kind of business from 30 states' revenue departments, linked state by state under "Check against states' sales-tax records" (used only as a check)
Minnesota Office of the State Auditor, Analysis of Municipal Liquor Store Operations, 2024 (sales of each city's municipal liquor stores, used to bound Minnesota's liquor stores)
City of Chicago: community areas, neighborhoods and 2020 tracts by community area
City of Chicago notice: “This site provides applications using data that has been modified for use from its original source, www.cityofchicago.org, the official website of the City of Chicago. The City of Chicago makes no claims as to the content, accuracy, timeliness, or completeness of any of the data provided at this site. The data provided at this site is subject to change at any time. It is understood that the data provided at this site is being used at one's own risk.”
This product uses the Census Bureau Data API but is not endorsed or certified by the Census Bureau.
City of Saint Paul notice (its TIF districts and vacant buildings): “The Geographic Information System (GIS) Data to which this notice is attached are made available pursuant to the Minnesota Government Data Practices Act (Minnesota Statutes Chapter 13). The GIS Data are provided to you “AS IS” and without any warranty as to their performance, merchantability, or fitness for any particular purpose. The GIS Data were developed by the City of Saint Paul for its own internal business purposes. The City of Saint Paul does not represent or warrant that the GIS Data or the data documentation are error-free, complete, current, or accurate. You are responsible for any consequences resulting from your use of the GIS Data or your reliance on the GIS Data. You should consult the data documentation for this particular GIS Data to determine the limitations of the GIS Data and the precision with which the GIS Data may depict distance, direction, location, or other geographic features. If you transmit or provide the GIS Data (or any portion of it) to another user, the GIS Data must include a copy of this disclaimer.”
City of Greeley notice (its TIF districts): the City of Greeley does not warrant its GIS data in any way and makes each of the disclaimers in its terms of use: the data is provided “as is” and “with all faults”, and the City assumes no responsibility or liability for its accuracy, completeness, timeliness or usefulness.
Local district and property layers under open licenses, credited as their licenses require: the Atlanta Regional Commission's community improvement districts, the Utah State Tax Commission's 2025 tax entities and the Utah Geospatial Resource Center's sales tax areas (CC BY 4.0); Sacramento County's property business improvement districts (CC0 public domain dedication); District of Columbia property (CC BY 4.0); Baltimore's vacant buildings (CC BY 3.0); the USDOT National Transit Map (CC BY 3.0 US); the City of Cleveland's land bank and its East and Shore to Core to Shore TIF districts (Open Database License: the parts of this site's data drawn from them are available under the same license); and DuPage County's TIF districts: Contains information licensed under the Open Geo-Spatial Data License – DuPage County (the license, in the layer's metadata). Every other district layer is named, with its publisher, where an area is selected.
Tax rules and rates: each state's revenue department, cited state by state; the Federation of Tax Administrators' 2017 survey of the sales taxation of services and its state rate table (January 2023), where a state's own rules haven't been read yet; and CaliforniaCityFinance.com (Coleman Advisory Services) for California's local tax measures and election results.

Neighborhood boundary credits

Who's On First (CC BY 4.0), compiled from: Atlanta Department of Planning and Community Development (CC BY 4.0); Baltimore Mayor's Office of Information Technology (CC BY 3.0); Washington Post neighborhood boundaries for Washington, DC (CC BY); Zolk Chicago Neighborhoods Map (CC BY 3.0 US); Azavea, Inc., Philadelphia neighborhoods (CC BY 3.0 US); Denver Community Planning and Development (CC BY 3.0); City of Cambridge GIS (CC BY); Quattroshapes, Foursquare (CC BY 2.0); Pediacities NYC neighborhoods (ODC-By 1.0); and, in the public domain or under CC0, Zetashapes, Mapzen, Mini Tenders, Burrito Justice, SanGIS/SANDAG, and the cities of Boston, Los Angeles, New Orleans, Oakland, Portland, San Francisco, San Jose, Santa Barbara, Seattle and St. Paul. Chicago community areas and neighborhoods: City of Chicago Data Portal. Boundaries are simplified and matched to census tracts here.

November 7, 2026.

You may look up any place, quote the site's figures with a citation and a link, and use them in your own planning, research and reports.

The site's estimates, its compilation of them and its reports and code are its own work, even where the figures behind them are public. Don't copy them in bulk, by script or otherwise, to republish, resell or build another product, without written permission from the site's owner; ask through "Suggestions and corrections" above.

The public sources the figures are built from stay under their own terms, listed under Sources in the method, and most can be had from their publishers. Data the site passes on under an open license (listed there) may be reused under that license: in the downloads, the City of Cleveland's land bank counts and its two TIF districts come from layers under the Open Database License and may be reused under it.

Figures are estimates, provided as they are, with no warranty that they are accurate, complete, current or fit for any purpose. Check them locally before any decision; using them is at your own risk. Some publishers ask that their own disclaimers be passed on; they are listed under Sources.

What the site records about visitors is described in the method, under "What this site keeps". These terms may change; the date above shows the current version.