Pay Now, or Pay More Later

Leaving people outside is not free. It shows up in the jail budget and at the hospital. It shows up when downtown businesses name homelessness as a top concern. We can name some of those costs, but we cannot total nor identify all of them.

This is a complex and divisive topic. No one wants homeless people sleeping on the streets downtown or in the nearby parks or woods any more than they want to be homeless. On that position, most can agree. Where the opinions tend to differ is how we pay for the solutions that end homelessness or at least greatly reduce the number of homeless people in our country. That is the point of this article.

We know what it costs to house someone in the local program that already does this work. We will never know the full cost of leaving the same person outside. The second number is always a guess, because most of what leaving someone outside costs never shows up on one bill.

If downtown felt safer, would more people eat there? Common sense says yes. More tables filled can mean more shifts, more sales tax, more visitors who come back. That is a fair question, but it is not a study. This article does not invent a tourism-loss number, because we found none in the county, city and state records that we could find.

Homeward Bound of WNC runs permanent supportive housing. That means a long-term apartment plus a case manager. Homeward Bound says regular meetings with a caseworker are required to stay in its supportive housing programs. Homeward Bound says this program serves "highly vulnerable, chronically homeless individuals."

The federal Continuum of Care grant for Homeward Bound's supportive-housing projects is $1,420,174 for the grant year that began in January 2026. That money covers scattered-site rentals and Key Commons. It does not cover the 85 units at Compass Point Village, which the organization told Asheville Watchdog are paid from another source.

Asheville Watchdog reported in November 2025 that the grant covers 102 households. That is about $13,923 per household in federal money for a year of rent, case management, and counseling. Homeward Bound adds a 25 percent match. Grant plus match is about $17,404 a household. Tenants who have income pay 30 percent of that income toward rent.

Homeward Bound's own site puts supportive housing at less than $15,000 a person a year. That figure counts people. Our $17,404 counts households. Asheville Watchdog describes those households as mostly individuals, with several families among them, and gives no count of people. The same page says a chronically homeless person's heavy use of public services costs $30,000 to $50,000 a year. The page gives no source for either figure. We found no Buncombe study that measures either one. The closest outside figure we found is a 2002 study of 4,679 homeless people with a severe mental illness in New York City, followed from 1989 to 1997. It put the public cost of one of them living on the street at $40,451 a year, in 1999 dollars. That lands inside Homeward Bound's range. It also counts a different group, in a different city, and its dollars are from 1999.

That is the "pay now" price we can put on paper: about $17,400 a household for the grant-funded program, plus whatever the tenant pays.

Some people living outside cost the public far more than others. The studies in this article that try to price "savings" mostly study that smaller group: people who have been homeless a long time, people with a disability, people with a mental illness, or people arrested again and again.

That is not most of the local count.

In February 2026, Buncombe counted 824 people on one night. 254 of them met the federal test for chronic homelessness. That is about three in ten of everyone counted. The test is narrower than the count. It reaches people in an emergency shelter, in a safe haven, or in a place not meant for human habitation, and it does not reach transitional housing. 245 people in the count were in transitional housing, so among the people the test can reach, the share is closer to four in ten. Broadly, the rule requires a disability, as federal homelessness law defines it, and about a year of homelessness in those places, with its own arithmetic for how the year is counted and for short stays in a jail or a treatment facility. The glossary has the full rule. In the same count, 91 adults reported a serious mental illness and 62 reported a substance use disorder. Answers in the count are self-reported.

The test also decides who can get some federally funded supportive-housing beds. A program can pledge beds to chronically homeless people, and say so in the records it files. When a tenant in one of those beds leaves the program, the next tenant has to meet the test too, unless no one in the local Continuum of Care's area does. Not every supportive-housing bed is pledged that way. Some federally funded projects use a wider door, where meeting the chronic test is one of six ways in. But for a pledged bed, a person usually has to have been homeless for at least a year in all, and have a disability.

Charlotte's Moore Place took in 85 chronically homeless adults. In the year before they moved in, the 61 tenants who released hospital records averaged 9.3 emergency-room visits each. At least one had 125. For those same years, the national survey put homeless people at about 1.4 visits a year. Those tenants were not the average person in a point-in-time count, the one-night PIT count.

So an average cost spread across all 824 people can be misleading. It is too high for most people in the count. It is much too low for a smaller group who cycle through the jail and the emergency room. This is but one reason why an average can only get us close, and why coming up with detailed specific costs is impossible.

Mental illness and addiction are part of why that smaller group is expensive. Inside Buncombe's jail, administrators say 42.6 percent of the people held have some kind of mental-health issue. That is the jail population, not the street count, and it does not mean those people were jailed because of mental illness, or that each one was in an active crisis. Sheriff Quentin Miller has said the jail is essentially the largest mental-health facility in Western North Carolina, and that the homeless situation and the mental-health crisis keep adding to the overcrowding. A jail major told county commissioners in June 2026 that the wait for a bed at Broughton Hospital in Morganton was six to 18 months.

The realistic split is this:

• Most people in the count are not in the "chronic plus disability" group.
• A smaller group uses crisis systems a lot. They cost more. Housing with support is built first for them.
• We should not talk as if every person counted costs what the highest-cost person costs.

For July 2026 through June 2027, the county's operating line for the jail is $26,429,357. That is the day-to-day cost of running it. Building work and the jail commissary sit on their own lines.

The jail was designed for 524 people. Staff say it can safely hold 470. As of August 2, 2026, it held 545 people on an average day, as reported by Mountain Xpress. On July 26 it held 548.

Spread the budget across 545 people, and the average cost of one jail bed is $48,494 a year. Use the year-to-date average WLOS reported in early August, about 505 people a day, and the same budget implies about $52,300. Either way, that is an average. It is not what the county pockets if one bed empties. A jail with fewer people still has to be staffed and run.

How much of the $26.4 million is homelessness? The county does keep a housing-status measure, and it released a one-page summary of it to Step Up AVL in September 2026. Asked for the records behind that page, it said there was nothing more it could share. A one-page county summary dated November 2025 defines "presumed unhoused" as "anyone with an address of 19 N Ann St. and 20 Davidson Dr. in the database." Those are the AHOPE Day Center and the jail itself. No one address can be both, so we read the county's "and" as either one. By that test, 1,288 of the 7,403 bookings in the year that ended June 2025 were presumed unhoused. The page gives that as 17 percent. In the year that began July 2025, up to a date the page does not give, it was 683 of 3,423, which the page gives as 19 percent.

Two things about that measure are worth knowing. It counts bookings, not people: the 1,288 bookings in the earlier year involved 696 people, so the same person was often booked more than once. And an address test is a rough one. Someone who gives a relative's address is not caught by it, and the county's own page says only that an address is often a more reliable indicator than the jail system's "homeless" flag.

A county analyst gave a different-sounding number in public. At a June 2, 2026, briefing, asked how many people leave the jail without housing, the analyst answered around 20 percent. That is the reporter's paraphrase, not a quotation, and it describes people leaving, which is not what the November 2025 page counts. The county has not said where the 20 percent came from. It released that page under our records request without saying which question the page answered, and when we asked where the page's figures came from, it replied that there is "no sharable document beyond what we have already provided."

The page stops there. It gives no stay lengths, and a dollar figure needs them, so the price below still comes from a national count. The Prison Policy Initiative read 175 jail rosters from July 2023 through June 2024. People recorded as unhoused were 4.5 percent of bookings, an undercount, the authors say, because people may give a shelter or a relative's address. Those people stayed 32 days on average, against 21 days for all bookings. The most serious charge they were booked on was most often trespassing. Because they stayed longer, we estimate they took up about 6.9 percent of jail nights. Against this jail budget, that is roughly $1.8 million a year, or about 37 people a day.

The two counts are not measured the same way. The national one uses whatever each jail's roster recorded. Buncombe's uses two addresses. Even so, the gap is wide. Unhoused people were 4.5 percent of bookings in the national sample. Presumed-unhoused people were 17 to 19 percent of bookings here, about four times the national share.

That does not make the local bill four times $1.8 million. The dollar figure depends on how long people stay, not only on how often they are booked, and the county produced no records of how long people stay here. What it does mean is that $1.8 million is a floor. For it to be the right figure, unhoused people would have to leave the Buncombe jail much faster than everyone else does, and nationally they do the opposite: they stay about half again as long. How much longer they stay here is the next thing we have asked for. Yet another example of why estimating these costs is difficult. But the jail bill is significant, and it is growing.

In a separate two-year sample, of the people recorded as unhoused who were booked and released inside the window, more than 40 percent were booked into the same jail system again within a year of that first booking. For people who were housed or whose housing was unknown, that share was 20 percent.

Thirty-seven fewer people a day would close about half the gap between the July 26 count and safe capacity. However, it would not save the county $1.8 million. The jail's medical contract is one example. Wellpath, the company that runs health care inside the jail, is paid a base amount that does not move while the jail's monthly average runs from 474 through 574 people. Above or below that band the county pays more or less, but only for the people outside it. Take 37 people off an average of 545 and the county owes Wellpath the same base amount.

We found nothing published by Mission Hospital or its owner, HCA, on patients' housing or what those patients cost. The figures below apply federal data and other states' rates to Buncombe's count. They are not a Mission invoice.

In 2020 and 2021, the CDC found that homeless people made 310 emergency-room visits a year per 100 people. Everyone else made 40. That's 675 percent more, almost eight times as many.

For how often homeless people are admitted to a hospital, the rates we used come from California and Massachusetts. Combined, treating one homeless person costs $2,978 to $7,381 a year more than treating a housed person. Across all 824 people in the 2026 count, that is $2.5 million to $6.1 million a year. The range is wide mainly because hospital-stay rates differ a lot by state. The notes at the end explain why that range could be too high or too low.

Forced psychiatric treatment is expensive too. Disability Rights North Carolina's 2025 review said a single day can cost well over $2,000, paid by public programs, insurers, and people with little or no coverage. It put the bill for one short episode, from the emergency room through a few days in a treatment facility, at $10,000 or more. We found no published figure on how many of those stays involve people without housing.

Add the cautious jail share to the hospital range and the pieces we could count come to $4.3 million to $7.9 million a year. That sum leaves out police, courts, shelter, detox, ambulance, involuntary psychiatric holds, and any school or child-welfare costs for the 67 children in the 2026 count.

A much higher ceiling comes from Denver. That trial studied people arrested at least eight times in three years. Public services used $25,544 a year on each person in the group not offered housing, across a wider set of costs. Part of that is what Medicaid was billed, not what it paid. Apply that figure to all 824 people here and you get about $21 million. That is a ceiling, not a local measurement. Denver did not study everyone the PIT count finds.

They almost do, if you ask the wrong question.

The wrong question is: "Does housing pay for itself?"

The studies that used a lottery to decide who got housing, in Denver and in Canada's Chez Soi, found the same basic pattern:

• People who get an apartment and a caseworker stay housed at high rates.
• They use some crisis services less.
• On the costs those trials actually counted, the drop covered about half of what the program cost. The other half of that cost is new money, and it buys the housing and the support.

Denver's difference was $6,876 a person a year in other public services. Chez Soi brought the net cost down by 46 percent from society's point of view, in 2016 Canadian dollars. Neither trial found that the counted savings covered the whole cost.

That is not a mixed signal. It is the same signal in two places:

• Housing costs money.
• Crisis use often decreases.
• The decrease we can count does not cover it all.
• The rest of the money helps buy a home.

Two research reviews tried to blend many studies into one ratio. One 2022 review landed near $1.80 in counted benefits per $1 of program cost in its middle result for American studies, and about $1.05, just above even, under its strictest test. A 2015 review of programs for people with mental illness said it doubted Housing First can be expected to pay for itself, and in the same breath said the programs still use resources better than traditional services.

Those ratios are why the evidence can feel like it argues with itself. A 1.80 and a 1.05 and a "does not pay for itself" cannot all be the headline. They are different studies, of different people, counting different bills, in different countries.

What they share is narrower, and more useful:

• They mostly studied the expensive minority, not the whole homeless population.
• They counted only some public bills: jail, shelter, emergency rooms, and the like. They did not count a quieter downtown, a restaurant that keeps its evening shift, or a visitor who does not cancel a trip.
• They cannot tell us what Buncombe would save next year if 37 jail beds emptied. A building still has to be staffed. A hospital still has to be open.

Homeward Bound reports that in 2025, 92 percent of people in its permanent supportive housing program stayed housed and did not return to homelessness. The organization does not say how many people that is. That is one program, one year. In Denver, the program's records run from 2016 through the end of 2020. That gave researchers three full years of follow-up on 243 people, the ones who had moved into the program's apartments before January 2018. Of those still alive after three years, 77 percent were in permanent housing.

The larger lesson is simple. We can name what it costs to house someone. We can never do more than make a broad estimate of what the same person would have cost in the jail, the hospital, the street, and the shops if left outside. Much of that second bill is impossible to total, and some of it is impossible even to guess.

What the studies settle is narrower than a saving, and more useful. The public is already paying. Right now that money buys jail nights and emergency-room visits, often for the same people more than once, along with police calls, court time and detox beds this article never priced. Housing and support shift part of that spending and bring part of it down. What those trials counted did not, on its own, cover the whole cost of the housing. Denver's researchers say their own figure is a conservative one and that the true offsets are likely to be greater than they were able to measure. So the choice is not whether to spend. It is what to spend it on.

Visitors spent $2.97 billion in Buncombe in 2023, $2.65 billion in 2024, and close to $2.6 billion in 2025, according to the yearly study prepared for Visit NC.

In April 2022, the Asheville Downtown Association presented a survey answered by 101 downtown businesses. The top answer on issues affecting business was reducing homelessness, followed closely by increasing safety.

Police calls about trespassing downtown rose from 474 in 2024 to 628 in 2025. The police data does not say how many of those calls involved homeless people.

We found no study that measures how many tourism dollars homelessness costs Buncombe. The 2025 visitor figure also covers the first full year after Hurricane Helene, so any effect of homelessness would be hard to separate from the storm.

One percent of $2.6 billion is $26 million, about the size of the jail budget. That is arithmetic, not a measured loss.

Could the hotel tax pay for housing? State law now limits that. Senate Bill 484 was signed June 22, 2026, and it covers every county and city in the state that charges a room occupancy tax. For proceeds collected on or after that date, it bars spending the tourism-related share on services a county ordinarily provides its residents, or on purposes designed for or mainly benefiting residents, unless a local act explicitly allows it. Affordable housing is one of the six barred uses it names.

So the downtown worry is on the record. The jobs-and-taxes story that may follow from a quieter downtown is easy to picture and impossible, with the sources we have, to price.

The $4.3 million to $7.9 million figure is a yearly estimate. Multiply it by the number of years, with nothing growing and no inflation:

• Three years: about $13 million to $24 million
• Five years: about $21 million to $39 million
• Ten years: about $43 million to $79 million

That is multiplication, not a forecast. It leaves out police, courts, shelter, detox, ambulance, forced treatment, and school or child-welfare costs for the children in the count.

The systems that absorb people when housing is missing are not standing still. The jail's average daily population rose about 55 percent from 2020 to 2026, from a year that included a pandemic drop in the pretrial population. The jail line went from $24.6 million adopted for July 2025 through June 2026 to $26.4 million for the year after. County public-safety spending went from $66.22 million in the budget for July 2016 through June 2017 to about $106 million in the budget for July 2026 through June 2027, not adjusted for inflation. That line covers more than the Sheriff's Office. The Wellpath medical contract took effect on July 1, 2026. Its base is $4,253,066 for the first twelve months. The term is two years, it renews for up to three more unless someone ends it, and the whole agreement depends on the county appropriating the money each year. By August 2026, booking cells were being used as temporary housing, and a second person in a one-person cell slept on a mat. The sheriff has talked about a behavioral-health wing and sending some people to other counties. The jail's operating line sits in the general fund, whose largest single source is the property tax.

Mission has told the state it is operating close to capacity. We found no figure for how much of that demand is homelessness.

The PIT count rose from 527 in 2021 to 824 in 2026. But the method was widened in 2024, and the 2025 count left out 1,548 people in FEMA hotels after Helene. We draw no growth rate from those counts. What they do show is that the 2026 count is the highest of the six.

Prevention is a third way to spend. Asheville-Buncombe is one of ten pilots for Right at Home: at least $5 million over three years for at least 1,000 households at risk of losing their housing. A Notre Dame study of the Santa Clara program that Right at Home grew from found that people who got that temporary help were 81 percent less likely to become homeless within six months and 73 percent less likely within a year. Until the national evaluation is published, no one can say how well the pilot works.

So, how do we pay for this, you ask? We already are. The only question is how much we spend and what we spend it on.