What a study can cover
One geography level, up to 50 areas, all in one state.
- A study covers a single level — state, county, place, tract, or block group — never a mix, so no one is counted twice.
- It can combine up to 50 areas, which may span several counties but must sit in one state, so every figure has one state to compare against.
- ACS is always included: it is the only source published at every level, and other sources' per-person figures divide by its population.
- Each person can keep 10 studies per workspace — their personal workspace, and each organization they belong to.
Why sources show different years
Each source in a study uses its own year, because each agency publishes on its own schedule.
- A new study starts with each source's latest release.
- A saved study keeps its years when newer data is loaded, so its figures never change by themselves. When a newer year has data for every area in the study, the study offers to switch to it, with an undo.
- Every figure is labeled with its source and year; a figure that combines sources lists each year it uses.
- A year can only be chosen when it has data for every area in the study. Tract boundaries changed in 2020, so older ACS years may not cover today's tracts.
- ACS figures are 5-year estimates: "2024" means survey responses collected from 2020 through 2024.
Where a source is unavailable
Some sources aren't available for every kind of area, so studies of those areas leave them out.
- CBP (County Business Patterns) is available for state and county areas only, so it is unavailable in place, tract, or block group studies.
- FDIC Summary of Deposits is available for county and tract areas only, so it is unavailable in state, place, or block group studies.
- SBA 7(a) and 504 loan data is available for county areas only, so it is unavailable in state, place, tract, or block group studies.
Which tracts are CDFI Investment Areas
The CDFI Fund marks a census tract an Investment Area when it meets any one of its economic-distress tests; GISBanker shows its list as published.
- A tract qualifies if it meets at least one test: a poverty rate of 20% or more; a median family income at or below 80% of the benchmark — the greater of its metro area's and the national metropolitan median, or for a tract outside metro areas, the greater of its state's and the national non-metropolitan median; an unemployment rate at least 1.5 times the national rate (5.4%); lying wholly inside an Empowerment Zone or Enterprise Community; or being in a county that lost at least 10% (metro) or 5% (non-metro) of its population between the 2011–2015 and 2016–2020 ACS.
- The current list was drawn from the 2016–2020 ACS on 2020 census tracts (published January 2023), so a tract's eligibility can disagree with its newer figures elsewhere in GISBanker.
- Investment Areas are one way a CDFI qualifies its Target Market; organizations can also qualify through the people they serve, which this list doesn't cover.
- Puerto Rico's tracts are in the CDFI Fund list but outside GISBanker's coverage. A handful of tracts can't be matched to today's tract map (Connecticut water tracts and Suffolk County, NY's retired 2020 codes) and are left out.
How the Demographics tab breaks down residents
Each share is a count from the ACS divided by the group it's a share of — for a study with several areas, the areas' counts are added up first, then divided.
- Shares are never averaged across areas: for a combined study every count is summed over its areas before dividing, so a large tract counts for more than a small one.
- Age, as a share of all residents (B01003_001E): under 18 (B01001_003E + B01001_027E + B01001_004E + B01001_028E + B01001_005E + B01001_029E + B01001_006E + B01001_030E); 18 to 24 (B01001_007E + B01001_031E + B01001_008E + B01001_032E + B01001_009E + B01001_033E + B01001_010E + B01001_034E); 25 to 44 (B01001_011E + B01001_035E + B01001_012E + B01001_036E + B01001_013E + B01001_037E + B01001_014E + B01001_038E); 45 to 64 (B01001_015E + B01001_039E + B01001_016E + B01001_040E + B01001_017E + B01001_041E + B01001_018E + B01001_042E + B01001_019E + B01001_043E); 65 and older (B01001_020E + B01001_044E + B01001_021E + B01001_045E + B01001_022E + B01001_046E + B01001_023E + B01001_047E + B01001_024E + B01001_048E + B01001_025E + B01001_049E).
- Sex, as a share of all residents (B01003_001E): female (B01001_026E); male (B01001_002E).
- Race, as a share of all residents (B02001_001E): white alone (B02001_002E); black or African American alone (B02001_003E); asian alone (B02001_005E); american Indian and Alaska Native alone (B02001_004E); native Hawaiian and Pacific Islander alone (B02001_006E); some other race alone (B02001_007E); two or more races (B02001_008E). Race and Hispanic origin are separate Census questions: people of any race can be Hispanic or Latino, so the two are shown apart and never added together.
- Hispanic or Latino origin, as a share of all residents (B01003_001E): hispanic or Latino (of any race) (B03002_012E). These don't cover everyone, so they don't add up to 100%.
- Education, as a share of residents 25 and older (B15003_001E): graduate or professional degree (B15003_023E + B15003_024E + B15003_025E); bachelor's degree (B15003_022E); associate's degree (B15003_021E). These don't cover everyone, so they don't add up to 100%. Shows completed college degrees. High school and some-college shares aren't shown yet: they need GED and some-college counts that aren't loaded.
- Getting to work, as a share of workers 16 and older (B08301_001E): drove alone (B08301_003E); public transit (B08301_010E); worked from home (B08301_021E). These don't cover everyone, so they don't add up to 100%.
- Where residents lived a year earlier, as a share of residents 1 year and older (B07001_001E): same house (B07001_017E); elsewhere in the same county (B07001_033E); another county, same state (B07001_049E); another state (B07001_065E); abroad (B07001_081E).
- Household income, as a share of households (B19001_001E): under $25,000 (B19001_002E + B19001_003E + B19001_004E + B19001_005E); $25,000 to $49,999 (B19001_006E + B19001_007E + B19001_008E + B19001_009E + B19001_010E); $50,000 to $74,999 (B19001_011E + B19001_012E); $75,000 to $99,999 (B19001_013E); $100,000 to $149,999 (B19001_014E + B19001_015E); $150,000 to $199,999 (B19001_016E); $200,000 or more (B19001_017E).
- A count shows as unavailable when the Census Bureau withheld any part of it; a real zero shows as 0.
- Median age is shown only for a single area: a median can't be added up across areas.
How BEA income and GDP are matched to areas
BEA publishes some counties only combined with a neighbor, so those figures cover the combination and are labelled that way.
- Personal income, per-capita personal income, population and GDP come from the Bureau of Economic Analysis for states and counties.
- BEA publishes 23 Virginia counties together with the independent cities inside or beside them (for example, Albemarle County with Charlottesville), and Maui County together with Kalawao County, Hawaii. For any of those, the figures are for the whole combination and say so; they are never presented as one county's own value.
- Per-capita personal income is the value BEA publishes, not recomputed. Real GDP is in chained dollars of the reference year BEA states, so it can be compared across years; current-dollar GDP is not adjusted for inflation.
- BEA revises roughly the last four years in every annual release, so the release date is shown with the figures.
How bank deposit markets are built
A bank's deposits in an area are the deposits booked at its branches there, as of June 30.
- Each branch is placed in a census tract from its coordinates, and in that tract's county. Its deposits count toward that tract and county only.
- A bank's market share is its branches' deposits divided by all branch deposits in the area.
- Market concentration (HHI) is shown for counties only. A single tract is too small to be a banking market, so tracts show deposits and shares without it.
- Branches FDIC lists in a county that could not be placed in a tract are left out of the totals, shares, and HHI, and their count and deposits are shown separately.
- Large booking offices — branches reporting more than $1 billion in deposits — are flagged. Banks book deposits gathered across a region or the whole country at such offices, so by default their deposits are left out of the totals, shares, HHI and trend, the same as in the Market Opportunity Score; you can count them instead. They always count as branches. The $1 billion line is a size-based estimate, not something the FDIC reports.
- Bank health (Texas Ratio, return on assets, capital) comes from each bank's June 30 call report for the same year, so it lines up with the deposit figures. It covers the whole bank, not just its branches in this area.
Which year other sources use
Sources you don't choose per study are included automatically, matched to the study's ACS year so every figure describes the same period.
- Time series — unemployment (BLS LAUS), employment and wages (BLS QCEW), personal income (BEA), gdp (BEA), fair market rents (HUD), and income limits (HUD) — use the same year as the study's ACS release, or the nearest year loaded when that year isn't available, and say so.
- Fixed designations — natural hazard risk (FEMA National Risk Index), rural-urban commuting area (USDA RUCA), and opportunity zones (Treasury designations) — use their current version, which applies to every year GISBanker offers.
- Changing a study's ACS year moves these sources with it. Sources you choose yourself (business patterns, deposits, loans) keep the years you set.
- Each is included only where the study's geography level has data.
How studies are compared
2 to 4 saved studies side by side, each at its own source years, with any years that differ flagged.
- Each study keeps the years it was saved with. For every source, the comparison shows the year each study uses and flags the source when those years differ, or when only some of the studies include it.
- "Align years" moves every study to the latest year all of them have data for, across all of their areas. It applies to the comparison only; the saved studies are not changed.
- Dollar figures are shown as published, in each year's dollars, and are not adjusted for inflation. When studies use different ACS years, compare incomes, home values and rents with that in mind.
- Studies at different geography levels (for example, tracts and a county) can be compared, and the comparison says so: a larger area averages over more neighborhoods, and some sources are not published at every level.
- Business patterns, deposits and small-business loans are added up across each study's areas. Business totals are all-or-nothing: if any county has no published total for the year, the study's totals are shown as unavailable rather than as a partial sum.
- Deposits are those booked at branches in the study's areas as of June 30, leaving out large booking offices (branches holding more than $1 billion), which are shown separately. An area with no branches counts as zero deposits.
- Small-business loans are SBA 7(a) and 504 approvals in the federal fiscal year (October through September), counted where the project is; an area with no approvals counts as zero.
How figures are combined for several areas
Counts are added up across the areas; median income, home value and rent are estimated from income, value and rent brackets.
- Counts — people, households, housing units, dollar totals — are added up, and rates are computed from those sums. If any area's count is suppressed, the combined count is shown as unavailable rather than as a partial total.
- Medians cannot be added or averaged. Median household income, home value and gross rent are estimated by adding up how many households fall in each bracket (for example $50,000 to $59,999) across the areas, then finding the bracket that holds the middle household and interpolating within it. This is the standard method, but it can differ slightly from what the Census Bureau would publish for the same area, because the Bureau uses finer internal data. Checked against the published medians of every U.S. county (ACS 2024), the estimate is typically within 1% for household income and within 0.1% for home value and rent. These figures are labelled as estimates.
- When the middle household falls in the highest, open-ended bracket, the median is shown as that bracket's lower bound with a plus sign, such as "$200,000+".
- Rent medians count only renters who pay cash rent, as the Census Bureau's published median does.
- Other medians (such as age and family income), per-capita income and the Gini index are published only for each area and are unavailable for a combined area.
- A study with a single area shows the published figures unchanged.
Which years each source covers
Most sources go back to 2018; a few start later or have only one current version, for the reasons below.
- ACS (American Community Survey): states and counties from the 2018 release; tracts from the 2020 release, the first on today's tract boundaries (earlier releases use the 2010 boundaries); places and block groups in the latest release.
- Tract income levels (low / moderate / middle / upper) start with the 2023 release. They compare each tract with its metro area as OMB defined metro areas in 2023, and earlier releases don't match those definitions.
- County Business Patterns from 2017, the first year on today's industry codes; FDIC, NCUA, SBA, BLS, BEA, HUD and FFIEC CRA from 2018.
- FEMA's National Risk Index has one current version: FEMA recalculates every area with each release, so versions aren't comparable over time.
- USDA rural-urban commuting codes are published once per decennial census; the 2020 codes are the current ones.
- Opportunity Zones have one designation round in effect so far (2018); designations don't change year to year.
How "what stands out" is decided
A figure stands out when the area is in the top or bottom 10% of comparable areas.
- Comparable areas are areas of the same kind: other states for a state; the other counties or places in its state; tracts or block groups in the same counties. A study of several areas is compared as one combined area.
- The percentile is the share of comparable areas with a lower value (ties count half). It's shown as, for example, "Higher than 94% of Texas counties". At least 10 comparable areas with data are needed.
- Percentiles are used instead of z-scores because incomes, home values and rates across areas are rarely bell-shaped; a percentile means the same thing whatever the shape.
- Standing out is not good or bad in itself — a high poverty rate and a high income both stand out.
How change over time is measured
Figures are compared with the ACS release 5 years earlier, with dollar amounts adjusted for inflation.
- Each ACS 5-year release pools five years of survey responses, so releases that overlap share most of their data. Following Census Bureau guidance, a figure is compared with the release 5 years earlier — for example 2024 (2020–2024) with 2019 (2015–2019) — whose years don't overlap.
- Dollar figures are restated in the later release’s dollars using the CPI-U-RS price index, as the Census Bureau recommends, so a change shows real growth rather than inflation.
- Counts are compared as a percentage change; rates as a change in percentage points.
- No change is shown when the earlier release has no data for every area in the study — for example tracts before 2020, which used the previous census’s boundaries.
How data is kept up to date
Every night GISBanker asks each publisher for its latest release and loads anything new, so new data usually appears within a day of publication.
- Only new or changed periods are loaded. Before and after each load, the rows for each period are compared, so the Data Sources page can show which periods arrived and which the publisher revised.
- ACS 5-year estimates: checks the latest release in the Census Bureau's API.
- County Business Patterns: checks the latest year in the Census Bureau's API.
- FDIC Summary of Deposits: checks the latest survey year in FDIC's API. Every 30 days the most recent year is reloaded, because the publisher revises figures after first publishing them.
- FDIC call reports: checks the latest quarter in FDIC's API. Every 30 days the most recent year is reloaded, because the publisher revises figures after first publishing them.
- NCUA credit union call reports: checks the newest quarter on NCUA's quarterly-data page.
- HMDA mortgage lending: checks the latest year CFPB's HMDA API serves. Every 30 days the most recent year is reloaded, because the publisher revises figures after first publishing them.
- BLS county unemployment (LAUS): checks the latest month in BLS's API. Every 30 days the most recent 2 years are reloaded, because the publisher revises figures after first publishing them.
- BLS employment and wages (QCEW): checks whether BLS has posted the next quarter's file. Every 30 days the most recent 2 years are reloaded, because the publisher revises figures after first publishing them.
- BEA personal income and GDP: checks the date BEA's county file was last republished (each release revises earlier years).
- SBA 7(a) and 504 loans: checks the extract date of SBA's newest loan file (each extract restates every year).
- HUD Fair Market Rents: checks the fiscal years whose HUD file GISBanker knows (HUD file names are irregular, so each new year is added by hand).
- HUD income limits: checks the fiscal years whose HUD file GISBanker knows (added by hand, as for Fair Market Rents).
- FEMA National Risk Index: checks whether FEMA serves a newer version of the index (each version recalculates every area).
- FFIEC CRA lending is loaded by hand: FFIEC blocks automated downloads, so its release can't be checked either. The check estimates when the next year should be out — a year after the previous year arrived, or November of the following year when that isn't known — and reminds the operator from then until it's loaded.
- USDA rural-urban commuting codes change once a decade, after each decennial census, and are loaded by hand.
- Opportunity Zones: the 2018 designations stay in effect until the next round takes effect on 2027-01-01; loading that round needs a loader update.
Why some Connecticut county data is missing
In 2022 Connecticut's 8 counties were replaced by 9 planning regions, so county figures published for the old counties can't be shown.
- GISBanker uses Connecticut's 9 planning regions as its counties, as the Census Bureau has since 2022.
- The old counties and the planning regions have different borders, so figures published for the old counties cannot be split or combined to fit the new ones. GISBanker shows them as unavailable rather than estimating them.
- Affected: County Business Patterns before 2022, BLS QCEW for 2018–2023, BEA personal income and GDP for 2018–2023, and any ACS release that reports Connecticut by its former counties.
- HUD Fair Market Rents and income limits are not affected: HUD publishes them by town, and every town maps to exactly one planning region.
How the Market Opportunity Score works
A 0–100 score of where demand is strong and banking supply is thin, ranked against comparable areas — 100 means more opportunity than every peer on every measure.
- Every measure is a rate, a per-resident figure or a growth rate — never a raw total — so small and large areas compare fairly. Each measure becomes a percentile among peers (the share of peers it beats, counting ties as half), turned around where lower is better, so a higher percentile always means more opportunity.
- Demand (30 points by default): population growth, higher is better; household growth, higher is better; median household income, higher is better; median household income growth, after inflation, higher is better.
- Business activity (20 points by default): business establishments per 1,000 residents, higher is better (state and county only); growth in business establishments, higher is better (state and county only); annual payroll per employee, higher is better (state and county only).
- Deposit capacity (20 points by default): bank deposits per resident, higher is better (state and county only); bank deposit growth, higher is better (state and county only).
- Underserved gap (20 points by default): bank branches per 10,000 households, lower is better; small-business loans per 100 business establishments, lower is better (state and county only); small-business loans per 1,000 households, lower is better (tract only).
- Risk (10 points by default): unemployment rate, lower is better; FEMA natural hazard risk score, lower is better (county and tract only).
- A component is the average of its measures' percentiles; the score is the weighted average of the components. You can change the weights; a component set to 0 is left out.
- Presets set the weights for a common question — balanced (demand 30, business activity 20, deposit capacity 20, underserved gap 20, risk 10), growth markets (demand 40, business activity 30, deposit capacity 10, underserved gap 10, risk 10), deposit gathering (demand 20, business activity 10, deposit capacity 45, underserved gap 15, risk 10), and underserved markets (demand 20, business activity 15, deposit capacity 10, underserved gap 45, risk 10). Weights needn't add up to 100: each component's share is its weight divided by the total.
- Peers are areas at the same level — counties are compared with counties — in the same state, or nationwide if you choose. A state with fewer than 20 areas at that level is compared nationwide automatically, and states are always compared nationwide.
- Not every source reaches every level. Tracts are scored on demand, underserved gap, and risk only, and the other components' weight is shared among those. Places and block groups show the individual measures but no score, because too few of them exist there for a score to mean much.
- A component needs at least half of its measures, and the score needs components carrying at least half of the weight; otherwise it is shown as unavailable rather than guessed.
- A study that combines several areas is scored as one area. Its counts (people, households, branches, loans, businesses, deposits) are added up across its areas and its median household income is estimated from income brackets, as for other combined figures; it is then ranked among single areas at the same level, with its own areas left out. FEMA's risk score is an index that can't be added up, so it is left out for a combined area and risk rests on unemployment. Adding areas together smooths out their extremes, so a combined area tends to land nearer the middle of the ranking than its most and least promising parts.
- Areas with fewer than 1,000 residents (counties and states) or 200 households (tracts) are not scored and are left out of everyone else's comparison: survey figures for very small areas swing too much to rank, and most such tracts are military bases, airports or parks.
- By default, deposit capacity leaves out large booking offices, and you can include them. The FDIC records deposits at the branch where a bank books them, never where depositors live, and large banks book deposits gathered across a region or the whole country at a few offices — so a county with one such office can look far wealthier than it is. A large booking office is any branch reporting more than $1 billion in deposits; the typical branch holds about $100 million. This is an estimate based on size, not something the FDIC reports. When they are left out, their deposits are excluded from both deposits per resident and deposit growth; they still count as branches. A few genuinely large city branches are left out too, which understates the largest cities slightly.
- Growth compares non-overlapping ACS releases (2019 and 2024) for states and counties. Tracts use 2020 and 2024, the earliest release on today's tract boundaries; those overlap by one survey year, which understates change. Dollar figures are adjusted for inflation with the CPI-U-RS before growth is measured.
- The score on each study's dashboard card uses the study's saved weights (or the defaults), ranks it within its state, and leaves large booking offices out — the score card's starting settings. Change them on the study's page.
- A high score is a reason to look closer, not a forecast: it is relative to the chosen peers, and it uses each source's latest year, which can differ between sources.
When the source data has a problem
Known problems in the published data are listed on the Data Sources page, and any figure they affect carries a note saying what happened.
- Agencies sometimes publish data with gaps or errors: a tract left out of one release, a total that can't be right, a county reported under codes that no longer exist. GISBanker never fills these in with a guess.
- Each known problem is handled one of four ways: not shown (the figures can't be placed in today's areas), shown as unavailable (the source has nothing usable), adjusted (matched to today's codes, with the change described), or shown with a caution (published as-is, but compare carefully).
- Every problem is listed on the Data Sources page with the areas and years it affects, and any figure for an affected area carries a note — including a note on a county or state when some areas inside it are affected.
- Each data load also records every record it had to skip or couldn't use, so any gap can be traced back to the source.