The Foundation for Job Creation — JobCreation.us
Research

One word, a hundred different situations

We are beginning the work of classifying homelessness in useful detail — because a veteran in a tent and a working mother in a car need entirely different things, and one word hides that.

745,652
people in America had no home on a single night in January 2025
HUD Annual Homelessness Assessment Report, 2025
266,320
of them slept outside, in vehicles, or in places not meant for sleeping
HUD, January 2025 Point-in-Time count
+81%
rise in chronic homelessness since 2013, with unsheltered homelessness up 36%
HUD, 2025 report

Why one number isn’t enough

Behind that single figure are people with almost nothing in common. A veteran in a tent. A family in a minivan with both parents working. A man who has been outside for eleven years and has a mental illness nobody is treating. A teenager who aged out of foster care last spring.

They get counted the same way, discussed the same way, and too often offered the same thing. It doesn’t fit any of them.

The car is the part that stopped us. Federal counts don’t require communities to track how many people are living in vehicles, so there is no national number. Where cities do count, the share is staggering — roughly 39% of the unsheltered population in Los Angeles County, and at points more than half in Oakland and Seattle.

Research on those households finds they look different from the stereotype: more likely to be working, retired, older, or female. A car is often the last thing standing between a person and the sidewalk.

What we intend to collect

Four dimensions, kept separate — because a person is usually several of these at once, and forcing one label destroys the information.

Where they sleep

Emergency shelter, transitional housing, a car, a van, an RV, a tent, the street, an abandoned building, a motel, or a couch in someone else’s apartment. The last two aren’t in the federal count at all, which is a large part of why school-district numbers run so much higher than HUD’s.

Who they are with

Alone, with a partner, with children, with other adults — and with animals. A dog is one of the most common reasons a person turns down a shelter bed, and almost nobody tracks it.

How long they have been out

First time, on and off, or long-term. The federal definition of chronic homelessness is twelve months, or four episodes in three years, alongside a disabling condition.

What they are carrying

Employment status. Veteran status. Age. Whether they are living with a substance use disorder, a serious mental illness, or a physical disability. Whether they recently left incarceration, foster care, or a violent household.

Does homelessness cost society?

Sometimes — and the answer is more useful once you break the population apart.

The cost is concentrated. A chronically homeless person runs roughly $30,000 to $50,000 a year in public expense, averaging about $35,578. But chronic cases are only around a fifth of the homeless population. The family sleeping in a Honda while both parents work costs the public close to nothing.

Where the money goes is medical, not moral. In one Arizona county study, a person with chronic mental illness who was homeless cost about $73,000 a year — and roughly 75% of that was health care, not police or jail.

And the intervention often costs less than the neglect. Central Florida found $31,065 per person per year while homeless against $10,051 in supportive housing. Denver measured $15,733 saved per person per year. Emergency department visits dropped by around 61% after two years housed.

These savings findings concentrate on the heaviest users of public services, and results vary by city and population. Some careful studies find housing improves stability without fully paying for itself. We will publish the range, not only the strongest numbers.

None of that is an argument that people are worth helping because they are expensive. It is an argument that the person in a car with a job is one broken transmission away from becoming the kind of case everyone agrees is a crisis — and that catching them there is both cheaper and more decent.

How we will collect it

  • With shelters, on whatever they are permitted and willing to share. Most client records sit in federally governed systems that cannot be handed to an outside organization, so we expect to work largely from aggregate counts.
  • Directly, from people living outside — asked in person, answered voluntarily, and only by people who identify themselves and why they’re asking.

We will never publish anyone’s name, face, or any detail that identifies them. The point is the pattern, not the person. Anyone may decline any question and still be treated like a human being.

This work is beginning

We are not publishing findings yet. We are saying plainly what we intend to do, so that shelters, outreach workers, and people living it can tell us where we’re wrong before we start.