Battle Creek, Michigan
Battle Creek
has jobs.
Why aren’t more of them held by people who live there?
Start with 100 people working in the city. Then look at where they call home.
An independent data story based on research from the W.E. Upjohn Institute and public Census data.
Imagine 100 people at work.
Loading the verified Census data…
Source: our calculation from Census LODES 8.4, 2021. Dots round to whole percentages, not individual people.
The interactive data could not load. Reload this page, or read the source report and methodology below.
A city line runs through
a regional labor market.
Battle Creek’s employers draw people from surrounding cities and townships. Its residents also find work across that same region.
Start with the jobs and you see people arriving. Start with residents and you see people leaving. Both views contain the same people who live and work in Battle Creek.
Follow the connection
Same city.
Two directions.
First follow city jobholders home. Then reverse the map to see where resident workers have jobs.
Where Battle Creek jobholders live
Private primary jobs, 2021
Circle area represents workers. Lines connect areas; they are not travel routes. Outlines show cities and townships.
Largest connections outside the city
Select a place to locate its connection.
Select a place to inspect the connection.
New geographic aggregation of Census LODES. Earnings are monthly bands in the job record, not household income. Download every connection (CSV).
The connection changes
with the paycheck.
Higher-earning jobs in the city draw an even larger share of workers from outside it. Yet many higher-earning residents also work elsewhere.
Compare the two questions below. The percentages count different populations, so they should never be added together.
Our calculations, Census LODES 2021, JT03. The upper band is more than $3,333 per month, the source behind the report’s approximate “$40K+” label. These are nominal earnings, not inflation-adjusted.
What the lines cannot tell us
A commute is a connection.
Access is a harder question.
A map can show where workers live and where their jobs are recorded. It cannot tell us who applied, who was turned away, which jobs are vacant, or whether a shift fits a parent’s day.
Commuting is a normal part of a regional economy. These counts do not show that incoming workers displaced residents. They also leave out residents without a job.
The question goes
beyond the city line.
A training program can open a door. Transportation and child care can make a shift possible. Pay and working conditions can make staying worthwhile.
Success would mean more residents able to reach work that improves their lives, wherever in the region that work happens to be.
Follow the evidence.
Independent interactive adaptation and analysis based on research from the W.E. Upjohn Institute and public data. This is not an official Upjohn product.
Original research
Iryna V. Lendel, Chloe Wieber, Sevrin Williams and Gerrit Anderson. Economic Outlook of Battle Creek: Challenges and Opportunities, Technical Report 25-054, September 2025. Repository publication: January 2026. Research highlight: February 17, 2026.
Barrier explanations summarize the report’s findings and recommendations. They are not causal effects estimated by this interactive. Report text, graphics and logos are not reproduced.
Public data
Census LEHD Origin-Destination Employment Statistics, 2021 private primary jobs (JT03), release 8.4, Michigan vintage 20251202_1657. OD main and auxiliary files, RAC and WAC totals, and the block geography crosswalk.
Census TIGERweb 2024 county subdivisions supply map outlines and representative points. ACS 2019–2023 estimates appear only as findings attributed to Upjohn, not as independently reproduced estimates.
What we counted and calculated
A primary job is a worker’s highest-paying job. This analysis keeps those primary jobs that are in the private sector. The LODES frame covers ages 14–99 and beginning-of-quarter employment in the second quarter. These are occupied jobs, not vacancies or a census of everyone who could work. Public-sector jobs, secondary jobs and self-employment are outside this analysis.
We select 2020 census blocks assigned to Battle Creek city (place GEOID 2605920) in the release’s 2024 geographic crosswalk. Sum workplace records for jobs in the city; sum residence records for resident workers. A record with both ends inside belongs to the shared “live and work here” group.
Inbound share = jobs in the city held by people living outside ÷ all covered jobs in the city. Outbound share = covered resident workers with jobs outside ÷ all covered resident workers. Each earnings view uses its own denominator.
The map groups Michigan blocks into county subdivisions. It retains every connection in the downloadable data and shows the largest regional connections as lines. Out-of-state residents of city jobholders are grouped directly from Michigan’s auxiliary file. For city residents, the difference between RAC totals and Michigan destinations is retained as an unlocated out-of-state / cross-file residual. We do not invent destinations for it.
Michigan workplace data for 2022 and 2023 are unavailable in this Census release. This is a historical view of 2021, not a live labor-market tracker. Definitions and coverage.
Why our totals differ from the report
The report identifies OnTheMap 2021 private primary jobs but does not provide the original export or release identifier. Our current-release sums do not match its exact counts. The table keeps both visible; the map and opening use the current-release calculation. We have not established the exact reason for the discrepancy.
| Measure | Upjohn table | Our calculation |
|---|
Validation: raw files checked against Census’s published SHA-256 checksums; OD workplace sums independently checked against WAC. Earnings bands and both geographic partitions sum to their respective totals. Exact replication of the report remains unresolved.
What this analysis cannot establish
LODES contains modeled and disclosure-protected administrative data. Small flows are not precise individual observations. Workplace addresses do not establish actual daily journeys or remote-work arrangements. Lines on this map are neither road routes nor travel times.
The data do not identify hiring discrimination, skill deficits, care needs, job vacancies or a causal effect of commuting. Different earnings groups may differ in age, industry, hours and other characteristics. We do not control for those differences.
Report statements about transit and child care refer to its historical sources. They should not be used as current service schedules. Proprietary Lightcast, Moody’s, Esri and REMI work cannot be fully reproduced from the public files acquired. The ACS data endpoint required an API key; its vehicle estimate is therefore attributed to the report. No values were inferred from chart pixels.
Research and data accessed October 5, 2026. Original analysis, writing and visualization. Reproducible Python scripts, source log and research notes accompany the project.