Here's what Daniel sent us — when we look at the homeless population in any major city, has anyone produced a rough estimate of the origin stories? How do people actually wind up on the street? Not the stereotypes, not the political talking points, but the data — what are the common pathways, and how do they vary by geography?
This is one of those questions where the answer completely changes depending on whether you're looking at point-in-time counts or longitudinal studies. And most public debate relies on the wrong one. Point-in-time is a single-night snapshot — it's like trying to understand a river by photographing one square meter of it. You'll see some water, maybe a fish, but you'll completely miss the currents, the eddies, where the water's coming from and where it's going.
The "one bad month" narrative versus whatever's actually happening. So the snapshot tells you who's homeless tonight. The longitudinal tells you who becomes homeless, for how long, and why.
The most comprehensive attempt to map these pathways came from a 2015 study by the University of Chicago's Chapin Hall research center. They tracked twenty-six thousand homeless individuals across fifty US cities and identified five distinct pathways. Not one story, not "people just make bad choices" — five different failure modes of different systems. And the methodology is important here because they weren't just doing intake surveys. They followed people for two years, so they could see the sequence of events, not just the moment of crisis.
That's the number to anchor on. What are they?
The first and largest is what they called the "housing crisis" pathway — thirty-five to forty percent of cases. These are people who were housed, often for years or decades, and then experienced an economic shock. Rent burden tips past fifty percent of income, a job loss hits, an eviction follows, and the median time from last stable housing to first homeless episode is six months.
That's not "one bad month." That's a slow-motion collapse.
And here's the detail that most coverage misses — forty to fifty percent of homeless individuals are employed. They have jobs. The wages just don't cover housing costs in most major cities. I want to sit with that number for a second, because it's genuinely startling. You can work full-time, sometimes more than full-time, and still not be able to afford a place to live. The 2024 HUD Annual Homeless Assessment Report counted six hundred fifty-three thousand people experiencing homelessness on a single night in January, up twelve percent from the year before. That's a labor market problem as much as a housing problem.
The largest group isn't the visible chronic homelessness people picture. It's people who were in apartments and then weren't. People who probably looked a lot like their neighbors until the math stopped working.
The second pathway is what researchers call "institutional discharge" — fifteen to twenty percent of cases. People leaving prisons, jails, foster care, psychiatric hospitals without any housing arranged. The 2023 Los Angeles Homeless Services Authority report found that forty-eight percent of homeless adults in LA County had a prior incarceration.
That's a pipeline, not a coincidence. And "discharge" is such a sterile word for what's happening — you're released from one institution directly into another category, which is "person with no place to go.
It's a pipeline. And it's not just prisons — youth aging out of foster care have something like a twenty-five percent chance of experiencing homelessness within four years. You turn eighteen, the state says "good luck," and there's no housing waiting. Think about what that transition looks like in practice. You've spent years in a system where adults make decisions about where you sleep, what you eat, when you go to school. Then on your eighteenth birthday, that structure vanishes. You might have a social worker who's trying to help, but the legal obligation ends, and the housing isn't there.
The state creates the housing, removes the person, and then releases them without housing. The institutional discharge pathway is basically the government creating its own homeless population.
I mean, that's the uncomfortable version of it, yes. Pathway three is relationship breakdown — twenty to twenty-five percent. Domestic violence, family conflict, divorce. The gender split here is stark: eighty percent of families entering shelter are headed by women fleeing abuse.
That's a statistic that should reshape how we think about shelter design entirely. If four out of five families in the system are women with children escaping violence, the entire intake process should be built around that reality. And yet most shelters weren't designed with that in mind — they were designed as generic emergency beds.
In many places it's not. The shelter system in most US cities was built around the model of single adult men, because that was the most visible homeless population decades ago. So you have families — women with kids — navigating a system that wasn't built for them, trying to find space in facilities that may not even allow children. Pathway four is the health crisis pathway — ten to fifteen percent. Serious mental illness, addiction, a disabling physical condition that destroys earning capacity and support networks. This is the group people picture when they think "homeless," but it's not the majority.
That's the key misconception, right? "Most homeless people are mentally ill." What's the actual number?
A 2023 study in the Journal of the American Medical Association found that about twenty-five percent of homeless individuals have severe mental illness. But seventy-five percent have experienced significant trauma. The trauma number is three times the mental illness number, and most policy debates invert that. We talk about mental illness as the cause, when trauma is far more prevalent and often precedes both the mental health challenges and the homelessness.
The framing should be "mostly traumatized, sometimes mentally ill" rather than the reverse. And trauma isn't a diagnosis you can medicate — it's a life experience that shapes how people respond to stress, to institutions, to offers of help.
Pathway five, the smallest at five to ten percent, is what researchers call the "chronic spiral" — multiple factors compounding over years, often starting with childhood trauma or foster care, then cascading through education disruption, early substance use, first arrest, and so on. These are the people who've been failed by multiple systems in sequence. By the time they're visible on the street, they've already passed through child welfare, special education, juvenile justice, maybe the VA — and each system treated them as someone else's problem.
Five pathways, and they're not mutually exclusive.
That's the crucial thing. Most people in the Chapin Hall data had two or three factors operating simultaneously. You lose your job, which strains your marriage, which triggers a health crisis, and then the eviction notice arrives. The pathways are categories of primary cause, but the lived experience is usually an intersection. It's more like a Venn diagram where the overlapping regions are where most people actually live.
How do researchers even reconstruct this? Asking someone on the street "how did you get here" seems like you'd get incomplete answers. Memory under stress isn't exactly reliable, and there's probably some amount of shame or defensiveness that shapes what people are willing to share with a researcher holding a clipboard.
They use something called retrospective life-event calendars. It's a structured interview technique where you walk backward month by month through the last two or three years — "Where were you living in March? Who were you living with? Were you working? Any hospital visits? Any contact with the justice system?" It builds a timeline that's more reliable than open-ended questions because it anchors memory to specific dates and events. The technique was originally developed for epidemiological studies — tracking disease progression — and it turns out to work remarkably well for housing trajectories too.
Like reconstructing a financial audit, but for someone's housing biography. You're going line by line through the ledger of someone's life and finding the moment where the balance tipped.
That's exactly the right analogy. And the timelines reveal something that point-in-time counts completely miss — most homelessness is episodic, not chronic. The California Policy Lab did a 2022 study showing the median homeless spell is seven months, but twenty-five percent of people exit within two weeks.
A quarter of people are out in two weeks? That's — I mean, that's barely enough time to even register in most systems.
These are people who experience a brief crisis — they sleep in a car for a few nights, they stay in a shelter for a week, and then they find a new situation. They never show up in the visible chronic homelessness statistics, but they cycle through the system. And here's why this matters so much for policy: the long-term homeless — the people you see on the street for years — are a small fraction, maybe fifteen to twenty percent of the total homeless population, but they consume the majority of shelter and emergency resources. So we're spending most of our money on the smallest group, while the episodic majority — the people who might be reachable with relatively light interventions — get almost nothing.
We're designing policy for the most visible group, who are actually the smallest group, and missing the episodic majority who might be reachable with much lighter interventions. The person sleeping in their car for a week doesn't look like a "homeless person" to most people, but they're in the data.
This is exactly the finding that led to the "housing first" approach. The logic is: if the common factor across all five pathways is lack of housing, then providing housing without preconditions should be the starting point. Finland has reduced homelessness by eighty percent since 2008 using this model — they simply give people apartments. No sobriety requirements, no medication compliance, no "prove you're ready." Housing first, then voluntary support services.
Eighty percent reduction. What did they start with? And how did they actually operationalize "just give people apartments" — because that sounds like the kind of thing that gets said in a policy paper and then gets completely tangled in implementation.
In 2008, Finland had about eighteen thousand homeless people. They converted shelters into permanent supportive housing units and built new housing specifically for this purpose. The key was that they stopped doing temporary shelter and started doing permanent homes. They literally converted shelter buildings into apartment buildings — same buildings, different legal status for the people living there. A shelter bed is a bed you don't own and can be kicked out of. An apartment is yours, with a lease, with tenant rights. By 2024, they were down to under four thousand, and most of those are people temporarily staying with friends or family — rough sleeping is nearly eliminated.
The Finnish approach is basically: the problem of not having a house is solved by giving someone a house. It's almost tautological when you say it out loud.
It sounds absurdly simple, but it works because it addresses the common denominator. Compare that to the US "treatment first" model, which typically requires sobriety or medication compliance before housing. The idea is that you need to address the underlying issues first. But the evidence keeps showing that people are more capable of addressing addiction or mental health when they have a stable place to live. The treatment-first model asks people to solve their hardest problems while living in the worst possible conditions for solving any problem.
It's hard to hold down a job or stay sober when you don't know where you're sleeping tonight. That's not a moral insight, it's a logistical one. Your cognitive bandwidth is consumed by survival. You can't plan a recovery program when you're trying to figure out where you'll be in six hours.
The cost argument flips the intuitive assumption on its head. Rapid rehousing — helping someone with first month's rent, a security deposit, and short-term case management — costs about twenty-five hundred dollars per household on average. An emergency shelter bed runs more than four thousand dollars per month. And that's before you count emergency room visits, police interactions, and incarceration costs that spike during homeless episodes. There was a study in Orlando that tracked the public costs of chronically homeless individuals and found they were averaging thirty-one thousand dollars per person per year in emergency services. Housing them cost about ten thousand.
The cheap option is actually the housing. The expensive option is the shelter. We've got the math exactly backward in most municipal budgets.
Houston demonstrated this at scale. Between 2011 and 2024, they reduced homelessness by sixty percent using a coordinated entry system and rapid rehousing. Coordinated entry means there's a single point of intake — you don't have to navigate seventeen different agencies, each with their own eligibility rules. One assessment, one housing plan, one path. They essentially built an operating system for homelessness response. Before coordinated entry, a homeless person in Houston might have to visit five different offices, fill out five different intake forms, and get five different answers about what help was available — or more likely, get told at each one that they didn't quite qualify for that particular program.
Houston is not exactly a progressive utopia. This happened under mayors from both parties. It wasn't ideological — it was pragmatic. They looked at the numbers and realized the current system was both more expensive and less effective.
The politics of this don't break cleanly along left-right lines once you look at the data. The "imported homelessness" myth — the idea that people move to cities specifically to access homeless services — is another example. The 2024 US Conference of Mayors report found that only fifteen to twenty percent of homeless individuals in a given city came from outside that metro area. Eighty percent or more are locals. They lost housing in the same community where they're now sleeping on the street.
That's a persistent myth in every city. "They all come here because we're generous." And the data says no, these are your neighbors who lost their apartment. The person on the corner probably went to the same grocery store as you six months ago.
There's a related misconception about employment. The idea that homeless people "just need to get a job" ignores that many of them have jobs. In Los Angeles County, a 2023 economic roundtable study found that a full-time minimum wage worker would need to work seventy-two hours a week to afford a one-bedroom apartment at fair market rent. That's not a behavioral problem, that's a math problem. You can have perfect attendance, a great attitude, and still be short eight hundred dollars on the first of the month.
Let me try to synthesize the geography question from the prompt. How do these pathways vary by country? The Chapin Hall study gave us a US map, but the prompt asked about variation by geography. Does the five-pathway framework hold up internationally?
The Chapin Hall five-pathway framework was validated primarily in US cities, but the proportions shift dramatically by context. In Japan, you have very low visible homelessness despite high poverty rates and a suicide rate that's among the highest in the developed world. The difference is several factors: stronger family networks that absorb housing crises before they become street homelessness, strict anti-loitering laws that make rough sleeping difficult, and a different social safety net structure. But Japan has massive hidden homelessness — people living in twenty-four-hour internet cafes and manga cafes, sometimes for years. The government estimated around four thousand people were living in internet cafes in Tokyo alone in a 2020 survey, and those numbers have likely grown.
It's not that Japan doesn't have housing crisis victims — they're just invisible to the official counts because they're in a cubicle with a computer rather than on a sidewalk. The pathway exists, but the terminal point looks different.
The UK presents another variation. Shelter, the housing charity, estimates over three hundred thousand people are rough sleeping or sofa surfing — what they call "hidden homeless." The pathways are broadly similar to the US, but the institutional discharge pathway is particularly pronounced because of changes to mental health bed availability over the last thirty years. The UK closed most of its long-stay psychiatric hospitals without creating sufficient community housing, and a significant portion of the street homeless population in British cities are people who would have been institutionalized a generation ago. It's the same institutional discharge pathway we see in the US, but the specific institution that's discharging people is different.
Different systems failing in different ways, but the pathways rhyme. The names of the institutions change, but the dynamic — person exits system, system has no housing plan, person ends up on the street — is remarkably consistent across borders.
Finland is the outlier that proves the rule. They looked at the same five pathways and said: what if we just solve the housing part for everyone, regardless of which pathway brought them here? And it worked. It's worth noting that Finland's approach is expensive upfront — they invested heavily in housing construction and conversion — but the long-term savings in emergency services, criminal justice, and healthcare more than offset it.
Let me push on something. The "housing first" evidence is strong, but I've heard a counterargument that it works better in some contexts than others — that Finland's success is partly attributable to a stronger pre-existing social safety net and a smaller population, and that scaling it to a place like Los Angeles is a different challenge. Is that a fair critique, or is it just an excuse for inaction?
That's a fair question, and the evidence is mixed. Housing first has shown strong results in multiple US cities — not just Houston, but also in veteran homelessness specifically. The US has reduced veteran homelessness by about fifty-five percent since 2010 using a housing-first approach combined with VA supportive services. That's a population of tens of thousands across all fifty states. So it does scale. But you're right that the surrounding infrastructure matters — housing first works better when there's also a functioning healthcare system, accessible addiction treatment, and employment support. It's not just "give someone keys and walk away." The keys are necessary but not sufficient.
It's "give someone keys and also have services available if they want them." The "if they want them" is the controversial part. That's where a lot of people get uncomfortable — the idea that you're providing housing without requiring participation in treatment.
The voluntary nature of the support services is what distinguishes housing first from older models. The traditional approach was essentially: prove you're ready for housing by demonstrating sobriety, medication compliance, and program participation. Housing first flips that — housing is the platform from which you might choose to address other issues, not the reward for having already addressed them. And the counterintuitive finding is that voluntary services actually get higher engagement than mandatory ones. When people aren't forced into treatment as a condition of keeping their housing, they're more likely to engage with it on their own terms.
Which makes intuitive sense if you think about Maslow's hierarchy. It's hard to work on your substance use disorder when you don't know where you'll sleep. The safety need has to be met before the growth needs become accessible.
There was a 2021 randomized controlled trial in Canada — the At Home/Chez Soi study — that tracked two thousand participants across five cities. Housing first participants spent seventy-three percent of their time stably housed over a two-year follow-up, compared to thirty-two percent for the treatment-as-usual group. And every dollar spent on the intervention generated about a dollar fifty in savings from reduced emergency service use.
It's not just morally preferable, it's fiscally net positive. You spend a dollar, you get a dollar fifty back. That's a better return than most government programs can claim.
In the Canadian context, yes. The US studies show similar patterns but the savings vary by local healthcare and policing costs. In cities with very high emergency room costs — which is most major US cities — the savings tend to be even larger.
Let's talk about the scarring effect you mentioned earlier. Homelessness itself becomes a cause of further homelessness — how does that work mechanically? What actually happens to someone during a homeless spell that makes the next spell more likely?
This is one of the most underappreciated findings in the literature. Once someone experiences a homeless spell, even a brief one, several things compound. First, employment: you lose your job or can't job-search effectively without an address, a place to shower, clean clothes. Even if you technically still have the job, showing up unshowered and exhausted after sleeping rough is going to get you fired eventually. Second, health: a homeless spell is physically damaging — exposure, poor nutrition, interrupted medication schedules — and people exit homelessness in worse health than they entered. Third, social connections fray. People are embarrassed, they lose touch with friends and family who might have helped, they burn through their support network. The couch you could crash on for a week in October isn't available anymore by January.
Even if the initial cause was purely economic — you lost your apartment because of a rent hike — after three months on the street, you now also have health problems, no job, and a thinner social network. The pathway multiplies. You entered through the housing crisis door, but by the time you've been out for a while, you've picked up features of the health crisis pathway and the relationship breakdown pathway too.
This is where the "one bad month" myth is harmful. It implies that if you just get through the crisis, you'll bounce back. The data shows that a homeless spell is not just a housing interruption — it's a compounding trauma that makes every subsequent challenge harder to solve. The California Policy Lab study found that people who experience homelessness for more than six months have dramatically worse outcomes across every metric — employment, health, family reunification — than people who are rehoused within the first month. The six-month mark seems to be a kind of inflection point where the secondary damage becomes self-reinforcing.
Speed of intervention isn't just about reducing suffering in the moment. It's about preventing the secondary damage that makes the problem harder to solve. Every week someone spends unhoused isn't just a week of misery — it's a week of accumulating disadvantages that will make rehousing harder and more expensive.
That's exactly the argument for rapid rehousing over emergency shelter. Shelter keeps you alive but doesn't stop the scarring. Rapid rehousing gets you back into stable housing quickly enough to preserve employment, health, and social connections. It's not just cheaper — it's more effective because it intervenes before the cascade. And this is where the shelter system has a perverse incentive: shelters are measured by how many beds they fill, not by how quickly they get people out of those beds and into permanent housing.
Which brings us to the open question that keeps researchers up at night. As housing costs continue to outpace wage growth — and we've seen this accelerate in most major cities — is the housing crisis pathway going to become dominant everywhere? Are the other four pathways going to be subsumed by the sheer economic pressure?
The trend lines are not reassuring. The HUD data showed a twelve percent increase in homelessness from 2023 to 2024, and the primary driver was the end of pandemic-era eviction moratoriums and rental assistance programs. Those programs were essentially a massive natural experiment in preventing the housing crisis pathway — and when they ended, homelessness spiked. The implication is that a large portion of the homeless population was being held at bay by temporary subsidies, and without them, the economic pathway asserts itself. We basically ran a nationwide test of "what happens if we prevent evictions" and the answer was "homelessness goes down." Then we stopped the test and homelessness went up.
We had a glimpse of what prevention looks like, and then we turned it off. We collected the data and then ignored it.
Now we're seeing the results. There's an interesting development on the predictive side, actually — Los Angeles County launched a homelessness prevention AI pilot in March. The idea is to identify at-risk households before they become homeless, using administrative data — eviction filings, utility shutoff notices, emergency room visits, school changes for children. The algorithm flags households that are showing the early warning signs of the housing crisis pathway, and caseworkers reach out proactively with rental assistance or mediation.
Rather than waiting for someone to show up at a shelter, you're looking for the pre-homelessness signal in the data exhaust of their life. The eviction filing that hasn't resulted in a court order yet, the utility bill that's sixty days past due, the kid who changed schools mid-year.
And the ethical questions here are significant — you're essentially building a vulnerability detection system that flags people without their knowledge or consent. But the alternative is waiting until they're on the street, at which point the intervention is more expensive and less effective. It's the kind of trade-off that public health systems have been making for decades with things like child protective services, but applied to housing.
I'm torn on that. On one hand, catching people before they fall is clearly better. On the other hand, the government building a "likely to become homeless" classifier feels like the premise of a dystopian novel. You can imagine a world where that flag doesn't trigger help — it triggers a pre-emptive eviction or a denial of services.
The implementation matters enormously. If the outreach is supportive — "we've identified that you may be at risk of housing instability, here are resources you can access voluntarily" — that's very different from punitive surveillance. But the same system that flags people for help could flag them for other purposes. The LA pilot is still early, and the safeguards will determine whether it's a model or a cautionary tale. The difference between "we're here to help" and "we're here to monitor" is entirely in the governance structure around the algorithm.
To answer the prompt directly — yes, researchers have mapped the origin stories. Five pathways, varying proportions by city and country, most people experiencing two or three simultaneously. The largest group is economic evictees, not the chronically homeless population that dominates public perception. And the solutions that work are pathway-specific — eviction prevention for the housing crisis group, transitional housing for institutional discharge, trauma-informed care for relationship breakdown — but all of them share the common requirement of actually providing housing.
The meta-insight is that our current data collection is fundamentally inadequate. Point-in-time counts are like measuring poverty by counting people on the street at noon on a Tuesday. They miss the episodic majority, they don't capture pathways, and they're highly sensitive to weather and police activity on the counting night. If it rains on the night of the count, you get a lower number — not because fewer people are homeless, but because more people found somewhere to hide from the rain. We need longitudinal tracking that follows individuals over time, which is what the Chapin Hall study did, but it's expensive and methodologically difficult.
Without that data, we're flying blind on policy. If you think homelessness is primarily caused by mental illness, you fund treatment programs. If you think it's primarily caused by housing costs, you fund rent subsidies. The pathway data suggests we need both, but in different proportions than most current budgets reflect. Right now we're funding based on the most visible story, not the most common one.
The origin stories matter because they determine which solutions we even consider. And the evidence keeps pointing to the same uncomfortable conclusion: the most effective intervention for homelessness is housing. Not as a reward for good behavior, not as the final step in a treatment program — as the first step.
Which sounds simple and is apparently very difficult to implement at scale, except in Finland, where they just did it. And in Houston, where they just did it. And at the VA, where they just did it for veterans. The pattern isn't mysterious — it's just politically difficult.
Finland also has a population of five and a half million and a parliamentary consensus that took years to build. But the principles are transferable. Houston transferred them. The VA transferred them for veterans. The question isn't whether it works — it's whether the political will exists to do it for everyone. And that's not really a research question anymore. The research is done. The randomized controlled trials have been run. We know what works. The remaining obstacles are not about evidence — they're about whether we're willing to act on it.
Now: Hilbert's daily fun fact.
Hilbert: In the high medieval period, Saharan salt merchants used stone slabs cut to standardized dimensions as a form of currency along the Niger River trade routes. A single slab weighing roughly thirty kilograms could purchase one adult male slave in Timbuktu, but the exchange rate varied by season because salt became more valuable during the rainy months when transport routes flooded. Ibn Battuta's travelogue from 1352 mentions a specific transaction where a merchant haggled for three days over the salt-to-gold ratio, complaining that the slabs had chipped edges and therefore constituted debased currency.
...salt futures trading in the fourteenth century. I have so many questions. Were the slabs stamped? Was there a central authority guaranteeing the weight, or was it purely reputation-based?
I don't know what to do with that information. But I'm now picturing a medieval merchant complaining about currency manipulation in exactly the same terms as a modern forex trader.
The question that keeps circling back for me is: as housing costs continue rising faster than wages, does the housing crisis pathway eventually absorb the other four? If rent burden is the common accelerator, then in a sufficiently expensive city, doesn't every pathway eventually route through economic eviction? The domestic violence survivor leaves her abuser and then can't afford rent. The person discharged from prison can't find a landlord who'll rent to them. The person with a health crisis can't work enough hours to cover the rent. The economic pressure becomes the final common pathway.
That's the structural question underneath all of this. The pathways are distinct at the point of origin, but they converge at the same failure point — you can't afford to live anywhere. And if that's the common mechanism, then the Finnish approach isn't just one option among many. It's the only one that addresses the convergence directly. Everything else is treating the symptoms upstream while the downstream failure point remains the same.
The other thing I keep thinking about is the scarring effect and what it implies for policy design. If the damage compounds with time, then the most cost-effective dollar you can spend is the one that prevents the first night on the street. Not the second month, not the shelter bed — the prevention dollar. And right now, our spending is almost entirely on the back end. We're spending billions on emergency rooms, shelters, and jails — all of which engage after the damage has started compounding.
Emergency rooms, jails, shelters — all downstream of the failure. We're paying for the consequences rather than the prevention, and paying more for worse outcomes. It's like we've built a healthcare system that only treats heart attacks but won't pay for blood pressure medication.
That's the episode, really. The data exists. The pathways are mapped. The interventions are tested. The cost-benefit analysis has been done, repeatedly, in multiple countries, and it keeps pointing to the same answer. What's missing isn't knowledge. It's the willingness to reorganize budgets around what the knowledge tells us.
This has been My Weird Prompts. Thanks to our producer, Hilbert Flumingtop, who apparently spends his free time reading fourteenth-century salt merchant ledgers and finding the exact medieval analogy for modern monetary policy.
If you want more episodes like this one, leave us a review wherever you listen — it helps other people find the show. We're at myweirdprompts.com and on all the podcast platforms. I'm Corn.
I'm Herman Poppleberry. See you next time.
