The New Business Map
Where Americans are starting businesses since 2019, and whether local incomes kept up.
Headline finding
Business applications per adult rose in 797 of America's 801 large counties between 2019 and 2025, and grew no faster where household incomes fell. If anything, they grew faster where incomes rose.
Nationally, applications for new Employer Identification Numbers rose from 13.7 to 21.0 per 1,000 adults, a 53% increase, and the rise reached almost every county with at least 50,000 adults. The post-2020 wave of new businesses was not concentrated where incomes were falling. Among counties where income change is measured precisely, formation grew faster where real household income rose. On this evidence it was an opportunity wave, not a hardship wave.
The typical county is the median. Only counties where the change in real household income is statistically clear are counted (falls must pass a false-discovery-rate test). Source: Censal Research analysis of US Census Bureau data.
Key stats
- US business applications rose 62% between 2019 and 2025, from 3.50 million to 5.67 million; per adult, they rose 53%, from 13.7 to 21.0 per 1,000 adults, according to Censal Research's analysis of Census Bureau data.
- Business applications per adult rose in 797 of the 801 US counties with at least 50,000 adults between 2019 and 2025.
- The median large US county saw business applications per adult rise 57% between 2019 and 2025.
- In large counties where income is measured precisely, business applications per adult grew a median 36% where real household income fell, and 59% where it rose by more than 10% (income change 2015–19 to 2020–24).
- Only 17 of 801 large US counties combined rising business applications with a statistically robust fall in real median household income, after correcting for testing every county at once.
- In Hinds County, Mississippi, real median household income fell 9.8% while business applications per adult rose 49%, one of the steepest income declines among large US counties.
- Among US metros with more than a million adults, Tampa–St. Petersburg had the fastest growth in business applications per adult (+86%, from 19 to 36 per 1,000) and New York the slowest (+22%) between 2019 and 2025.
- Sheridan County, Wyoming received 1,806 business applications per 1,000 adults in 2025, 1.8 per resident adult, because registered-agent firms there file for clients nationwide.
Source: Censal Research, The New Business Map (2026), analysis of U.S. Census Bureau Business Formation Statistics, American Community Survey and Population Estimates.
1. The rise is nearly universal
Of 801 counties with at least 50,000 adults (about 83% of the US adult population), 797 recorded more business applications per adult in 2025 than in 2019. The median county rose 57%. The four exceptions are small movements: Burleigh ND (−16%), Santa Fe NM (−4%), Kent RI (−3%) and Cumberland ME (−1%). Among the 39 ranked metros with more than a million adults, Tampa (+86%), Seattle (+79%), Louisville (+77%), Columbus (+65%) and Phoenix (+65%) grew fastest per adult; New York (+22%), Washington (+26%), Atlanta (+27%) and Boston (+30%) grew slowest.
Metros containing a formation-service hub are not ranked. Source: Censal Research analysis of US Census Bureau data.
Show the numbers as a table
| Metro area | Per 1,000 adults, 2019 | 2025 | Change |
|---|---|---|---|
| Tampa-St. Petersburg-Clearwater, FL | 19.1 | 35.6 | +86.2% |
| Seattle-Tacoma-Bellevue, WA | 13.1 | 23.5 | +79.3% |
| Louisville/Jefferson County, KY-IN | 11.9 | 21.1 | +76.5% |
| Columbus, OH | 13.8 | 22.8 | +65.0% |
| Phoenix-Mesa-Chandler, AZ | 15.7 | 25.9 | +64.9% |
| Raleigh-Cary, NC | 17.0 | 27.4 | +61.0% |
| Portland-Vancouver-Hillsboro, OR-WA | 13.2 | 21.1 | +60.1% |
| Denver-Aurora-Centennial, CO | 21.0 | 33.4 | +59.0% |
| Memphis, TN-MS-AR | 14.7 | 23.1 | +57.6% |
| Cincinnati, OH-KY-IN | 11.2 | 17.5 | +56.5% |
| Riverside-San Bernardino-Ontario, CA | 9.8 | 15.3 | +55.6% |
| Indianapolis-Carmel-Greenwood, IN | 15.2 | 23.6 | +55.5% |
| Houston-Pasadena-The Woodlands, TX | 16.5 | 25.2 | +52.7% |
| Pittsburgh, PA | 8.9 | 13.5 | +52.3% |
| St. Louis, MO-IL | 13.3 | 20.1 | +51.5% |
| Milwaukee-Waukesha, WI | 12.7 | 19.1 | +50.9% |
| San Antonio-New Braunfels, TX | 12.0 | 18.1 | +50.7% |
| San Francisco-Oakland-Fremont, CA | 12.4 | 18.6 | +50.2% |
| Dallas-Fort Worth-Arlington, TX | 17.5 | 26.2 | +49.4% |
| Nashville-Davidson--Murfreesboro--Franklin, TN | 15.2 | 22.6 | +49.2% |
| Detroit-Warren-Dearborn, MI | 16.4 | 24.3 | +48.3% |
| Chicago-Naperville-Elgin, IL-IN | 14.2 | 21.1 | +48.1% |
| San Diego-Chula Vista-Carlsbad, CA | 11.7 | 17.3 | +47.9% |
| Virginia Beach-Chesapeake-Norfolk, VA-NC | 13.9 | 20.5 | +47.8% |
| Charlotte-Concord-Gastonia, NC-SC | 16.7 | 24.6 | +47.0% |
| Minneapolis-St. Paul-Bloomington, MN-WI | 12.5 | 18.3 | +46.7% |
| San Jose-Sunnyvale-Santa Clara, CA | 10.5 | 15.3 | +45.6% |
| Orlando-Kissimmee-Sanford, FL | 23.9 | 34.7 | +45.1% |
| Miami-Fort Lauderdale-West Palm Beach, FL | 34.7 | 49.7 | +43.2% |
| Las Vegas-Henderson-North Las Vegas, NV | 19.2 | 27.2 | +41.6% |
| Jacksonville, FL | 19.6 | 26.9 | +37.5% |
| Los Angeles-Long Beach-Anaheim, CA | 15.9 | 21.5 | +34.9% |
| Providence-Warwick, RI-MA | 9.1 | 12.0 | +31.5% |
| Oklahoma City, OK | 16.5 | 21.7 | +31.2% |
| Baltimore-Columbia-Towson, MD | 15.2 | 19.9 | +30.9% |
| Boston-Cambridge-Newton, MA-NH | 11.4 | 14.8 | +29.7% |
| Atlanta-Sandy Springs-Roswell, GA | 28.3 | 36.0 | +27.1% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 18.0 | 22.7 | +25.9% |
| New York-Newark-Jersey City, NY-NJ | 18.1 | 22.1 | +22.3% |
2. Falling incomes did not drive it
Real median household income (2024 dollars) rose in 637 of the 801 counties on point estimates, and the rise is statistically significant in 358. Across all large counties, median growth in applications per adult was 55% where real income fell by more than 2.5% and 62% where it rose by more than 10%, with no pattern in between. That flat picture understates the relationship, because survey estimates of income change in smaller counties are mostly noise.
Where income change is measured precisely (margin of error under 3% of income, 197 counties), median formation growth climbs from 36% where income fell to 59% where it rose by more than 10%. Among the 223 counties with at least 250,000 adults, the correlation between income change and formation growth is 0.24 (95% interval 0.11 to 0.37). The association is modest and descriptive, but it points the opposite way from the "necessity entrepreneur" story.
Counties where the margin of error on the income change is under 3% of income. The pattern is an association, not proof of cause. Source: Censal Research analysis of US Census Bureau data.
3. Seventeen large counties combine rising applications with a real income decline
Testing 801 counties at once produces false alarms: at 90% confidence about 40 counties would show a "significant" income decline by chance alone, and 38 do. We therefore apply a false-discovery-rate correction (Benjamini–Hochberg, 10%) across all counties. 17 counties pass, home to 4.1 million adults; 11 pass the stricter 5% bar.
They are mostly mid-sized Southern and Midwestern counties and outer suburbs: Hinds MS (Jackson, −9.8%), San Juan NM (Farmington, −9.9%), DeKalb and Peoria IL, Tippecanoe IN, East Baton Rouge LA, Fort Bend TX, Richland SC and Prince George's and Baltimore counties MD among them. Even here formation did not grow unusually fast: the median county in this group rose 49%, against 57% where incomes rose significantly.
| # | County | Adults (2025) | Applications per 1,000 adults, 2019 → 2025 | Real median household income, 2015–19 → 2020–24 (2024 $) | Income change ± margin of error | Passes 5% test |
|---|---|---|---|---|---|---|
| 1 | Prince George's, MD | 767,504 | 23.9 → 26.3 (+10%) | $104,196 → $101,798 | −$2,398 ± $1,752 | No |
| 2 | Fort Bend, TX | 725,439 | 19.4 → 28.2 (+45%) | $119,930 → $114,041 | −$5,889 ± $3,138 | Yes |
| 3 | Baltimore County, MD | 669,461 | 16.5 → 20.7 (+25%) | $94,314 → $91,768 | −$2,546 ± $1,853 | No |
| 4 | East Baton Rouge, LA | 356,418 | 21.1 → 25.2 (+19%) | $67,421 → $63,071 | −$4,350 ± $2,704 | Yes |
| 5 | Richland, SC | 349,502 | 16.8 → 24.8 (+48%) | $67,199 → $63,784 | −$3,415 ± $2,085 | Yes |
| 6 | Henry, GA | 200,706 | 31.3 → 39.0 (+24%) | $87,470 → $83,146 | −$4,324 ± $3,156 | No |
| 7 | Hinds, MS | 165,307 | 22.8 → 34.0 (+49%) | $54,755 → $49,402 | −$5,353 ± $2,224 | Yes |
| 8 | Tippecanoe, IN | 152,449 | 7.2 → 11.9 (+65%) | $65,190 → $60,636 | −$4,554 ± $2,426 | Yes |
| 9 | Peoria, IL | 137,477 | 8.2 → 14.8 (+81%) | $68,518 → $65,108 | −$3,410 ± $2,219 | Yes |
| 10 | Ector, TX | 123,393 | 15.2 → 24.5 (+62%) | $77,267 → $71,536 | −$5,731 ± $4,363 | No |
| 11 | Comanche, OK | 93,893 | 8.7 → 13.2 (+52%) | $64,001 → $60,761 | −$3,240 ± $2,352 | No |
| 12 | San Juan, NM | 91,500 | 6.4 → 13.3 (+106%) | $61,985 → $55,872 | −$6,113 ± $2,540 | Yes |
| 13 | DeKalb, IL | 80,297 | 5.9 → 11.8 (+99%) | $77,689 → $70,724 | −$6,965 ± $3,485 | Yes |
| 14 | Ozaukee, WI | 75,358 | 10.2 → 14.4 (+41%) | $104,558 → $96,996 | −$7,562 ± $4,527 | Yes |
| 15 | Belmont, OH | 52,684 | 5.5 → 10.4 (+90%) | $62,459 → $57,017 | −$5,442 ± $3,751 | Yes |
| 16 | Ward, ND | 52,197 | 9.7 → 11.9 (+23%) | $84,504 → $78,238 | −$6,266 ± $4,772 | No |
| 17 | Harrison, WV | 51,022 | 7.5 → 12.3 (+63%) | $65,058 → $60,377 | −$4,681 ± $2,946 | Yes |
Sorted by adult population. "Passes 5% test" shows whether the county also passes the stricter 5% false-discovery-rate bar.
4. Ten metros show the same pattern
Among 355 ranked metropolitan areas, applications per adult rose in 353. In 10, real median income fell by an amount that survives the same correction. Louisiana accounts for three of the ten.
- Farmington, NM (−9.9%)
- Odessa, TX (−7.4%)
- Minot, ND (−7.2%)
- Wheeling, WV-OH (−7.1%)
- Lafayette–West Lafayette, IN (−5.2%)
- Lawton, OK (−5.1%)
- Lafayette, LA (−4.5%)
- Jackson, MS (−4.3%)
- New Orleans, LA (−3.8%)
- Baton Rouge, LA (−3.7%)
Change in real median household income, 2015–19 to 2020–24.
5. Self-employment barely moved
The self-employed share of the employed rose by 0.5 percentage points in the median county; the change is significantly positive in 205 counties and significantly negative in 37. It is unrelated to application growth (correlation 0.05). Business applications are a leading, noisy indicator; the self-employment data through 2024 do not yet show the boom converting into a larger self-employed workforce.
National context only, not a finding of the study. Source: US Census Bureau, Business Formation Statistics.
Where formation grew fastest, and why the top of that list needs care
The fastest-growing large counties per adult are Sangamon IL (Springfield, +236%), Spokane WA (+182%), Stark OH (Canton, +173%), Vermilion IL (+166%) and Pinellas FL (+149%). The first three sit close to our formation-service screen: their 2025 rates are 1.4–1.5 times their state's rate. Applications are recorded at the address on the form, and registered-agent and company-formation firms tend to cluster in particular cities, often state capitals. We cannot separate that activity from genuine local start-ups in public data, so treat individual top-of-table rankings as indicative. The study's conclusions rest on the near-universal rise and the income comparison, not on any single county's rank.
A note on America's LLC mills
Some counties receive business applications from people who do not live there. Sheridan County, Wyoming, with about 26,000 adults, logged 47,787 applications in 2025, 1.8 per adult, because registered-agent firms there file for clients nationwide. Laramie and Natrona counties WY, Kent County DE and Flathead County MT (home to Montana's vehicle-registration LLC industry) show the same pattern at 96–184 per 1,000 adults. Wyoming's statewide rate quadrupled to 171 per 1,000 adults. These counties, and 35 others that meet a published three-part rule, are reported separately and excluded from every ranking above.
How we did it, in plain English
Read the full methodologyWe counted every application for an Employer Identification Number filed from each county in 2019 and 2025 (Census Bureau Business Formation Statistics, county file released June 2026) and divided by the adult population (Census Population Estimates). We compared the change with the change in real median household income between the 2015–19 and 2020–24 American Community Survey five-year estimates, converted to 2024 dollars with CPI-U, and with the self-employed share of workers from the same surveys.
Because ACS figures are survey estimates, we carried their margins of error through every calculation, corrected for testing hundreds of counties at once, and checked the income relationship in the counties where it is measured precisely. Counties with fewer than 50,000 adults were excluded to keep estimates stable; Connecticut was excluded because its counties were redrawn in 2022; formation-service hubs were screened out by rule. Every headline count was recomputed by a second, independently written program.
Limitations
- Associational only. This compares two periods across counties. It cannot say that income changes cause business formation or the reverse.
- Applications are not businesses. An EIN application signals intent; most applications never become employer businesses.
- Survey error. ACS county medians carry margins of error of roughly 2–6%. We test rather than ignore them; 17 counties pass the multiple-testing correction at 10%, 11 at 5%.
- Geographic attribution. Applications are located by the address on the form; registered agents and virtual offices concentrate counts in some counties. Our screen removes the clear cases; counties just below its thresholds (Sangamon IL, Spokane WA, Stark OH) remain in the tables and are flagged as cautions.
- Timing. The ACS periods are centred on 2017 and 2022, so income changes are smoothed and lag the 2025 application data.
- Disclosure noise. County counts carry differentially private noise (typical error about ±2 applications), immaterial above our threshold.
- Threshold. Results for counties under 50,000 adults are not reported.
Questions people ask
Are more Americans starting businesses since the pandemic?
Yes. Applications for new businesses (EIN applications) rose from 13.7 to 21.0 per 1,000 adults between 2019 and 2025, a 53% increase. The rise reached 797 of the 801 US counties with at least 50,000 adults.
Are people starting businesses because their income fell?
Not on the county evidence. Censal Research found that business applications grew no faster in counties where real household income fell than in counties where it rose. Where income change is measured precisely, applications grew a median 36% in counties where income fell and 59% in counties where it rose by more than 10%. The association is modest and descriptive, not causal.
Which US counties had the fastest growth in new business applications?
Among counties with at least 50,000 adults, the fastest growth per adult between 2019 and 2025 was in Sangamon County IL (+236%), Spokane County WA (+182%), Stark County OH (+173%), Vermilion County IL (+166%) and Pinellas County FL (+149%). The first three have rates well above their state's, which may partly reflect company-formation services filing from those counties.
Which big US cities had the fastest growth in business formation since 2019?
Among metro areas with more than a million adults, Tampa–St. Petersburg (+86%), Seattle (+79%), Louisville (+77%), Columbus OH (+65%) and Phoenix (+65%) grew fastest in business applications per adult. New York (+22%), Washington (+26%), Atlanta (+27%) and Boston (+30%) grew slowest.
Where did household incomes fall while new businesses rose?
Seventeen large US counties combine rising business applications with a fall in real median household income that survives strict statistical testing. They include Hinds County MS (−9.8%), San Juan County NM (−9.9%), DeKalb County IL (−9.0%), Tippecanoe County IN (−7.0%), East Baton Rouge Parish LA (−6.5%), Fort Bend County TX (−4.9%), Prince George's County MD (−2.3%) and Baltimore County MD (−2.7%). Ten metros show the same pattern, including New Orleans, Baton Rouge and Jackson MS.
Why does Wyoming have so many business applications?
Because registered-agent firms in Wyoming file company paperwork for clients across the country. Sheridan County WY recorded 47,787 business applications in 2025 for about 26,000 resident adults (1,806 per 1,000 adults), and Wyoming's statewide rate rose from 42 to 171 per 1,000 adults since 2019. Censal Research excludes such counties from its rankings.
How many business applications are filed in the US each year?
About 5.67 million in 2025, up from about 3.50 million in 2019, according to Census Bureau Business Formation Statistics. These are annual totals of the unadjusted monthly figures, which is how Census reports annual counts; adding up the seasonally adjusted months instead gives about 5.64 million for 2025.
Do business applications mean new businesses?
Not necessarily. A business application is a request for an Employer Identification Number; it signals intent. Most applications never become businesses with employees.
How to cite this study
Censal Research (2026). The New Business Map: Where Americans are starting businesses since 2019, and whether local incomes kept up. Version 1.1, October 2026. https://censalresearch.com/studies/new-business-map/
Short form for journalists: "according to an analysis of Census Bureau data by Censal Research".
Derived tables are published under CC BY 4.0; please credit "Censal Research, The New Business Map (2026)".
Data
- County table (CSV, 801 rows)Every ranked county: applications per adult in 2019 and 2025, real income change with margin of error, significance tests, false-discovery-rate q-values and flags.
- Metro table (CSV, 355 rows)Every ranked metro area, with the same measures.
- MethodologySources, definitions, tests, exclusion rules and the corrections log.
Journalists and researchers can ask for fuller tables, the analysis code, or a cut of the data for their area at [email protected].
All inputs are public U.S. Government statistical products. No restricted microdata was used. This product uses the Census Bureau Data API but is not endorsed or certified by the Census Bureau.
Author: Censal Research. Contact: [email protected]. Corrections policy: errors are corrected in place with a dated note.
Version history: v1.1 (October 2026) adds multiple-testing correction and a measurement-error check on the income relationship, and tightens the formation-service screen; v1.0 (September 2026) was not published.