Goldilocks in Skid Row (Act III)

The return of Goldilocks

quandt_007

Note that the post below has visuals that may not appear in your inbox.

We’ve toured our metaphoric Skid Row. First stop, the front gate, where we saw that a “free” bed isn’t free if the queue is endless or the porch swarms with pick-pockets; Second, the bunk-room, where safety rules, treatment pain, and crime math decide whether Joe stays or bolts. Lesson learned? Capacity, carrots, and sticks move as a bundle. Nudge one knob and the others shift, sometimes enough to shove people out of the just opened door.

Time to taste the porridge.

If our models have bite, they need to do two things:

1. Mirror what’s already happened on and off the streets;

2. Predict what’s likely to happen next.

That’s hard when the best data is locked in filing cabinets. Service providers rarely share detailed numbers.¹ So we’ll come at our models obliquely and—fair warning—tentatively.

Let’s keep to the fore George Box’s dictum, too: “All models are wrong, but some are useful.” We simplify to find levers policymakers can pull. Better that than letting those levers swing with the political headwinds—better, still, if we can explain why a program hits gold or face plants.

Goldilocks Catch-22

A high-level prediction of our models is that Treatment First and Housing First programs, alone, will fail to reduce homelessness. The former has long forms, sobriety checks, curfews, and mandatory meetings—hoops one must jump through to gain access to tiered shelters. The entry cost is so high for most that the most at-risk people fail. Housing First lives at the other extreme: walk in tonight, no questions asked. Low barriers may pull everyone in—including folks with serious untreated issues—which swells counts, raises theft risk, and can turn a quiet wing into a small, nervy city.

Too many rules lock people out; too few rules lets crowding and victimization hollow the place out from within. Not to mention that if the latter reduces treatment uptake, the issues that keep people from autonomous living persist.

The “just-right” mix requires a package deal:

1. Enough capacity to shrink queue’s plus

2. Visible safety guarantees—on-sit staff, clear, enforced norms, quick consequences—to keep the risk gap small plus

3. Real treatment incentives that make showing up for counseling or taking meds the easy choice.

Bus passes and food vouchers nudge some folks, but stronger levers—conditional cash, step-down privileges, time-limited rent—often matter more. This is the tension homelessness policy must manage: keeping the barrier at the right height, which likely differs across persons.

The move from “earn it” to “it’s yours”

Federal help once followed a strict staircase in the programs it funded to address homelessness. The client began with an emergency cot for the next, then spent months in a transitional program with curfews, sobriety checks, therapy sessions, and job-readiness classes. Only after they proved “housing-readiness” did they reach the top step—subsidized permanent housing. The U.S. Department of Housing and Urban Development (HUD) formalized this “Continuum of Care” in the mid-1990s; communities had to show a neat staircase to win McKinney-Vento dollars.²

But the staircase soon wobbled. Some street outreach workers complained the stairs for their clients were too steep. Requiring sobriety or perfect program compliance left the high risk under the bridges and on park benches. Psychologist Sam Tsemberis put data behind the gripe. In New York, he randomly placed some chronically homeless with mental illness in regular apartments with no preconditions and others in the old staircase. Two years later, 80 percent of his “Housing First” group were still housed versus roughly one-third of the control group, and their mental health or substance abuse outcomes were no worse.

Pilot projects in Los Angeles, Seattle, Toronto, Denver, and Philadelphia echoed those results and showed governments were spending a fortune cycling a small group through shelters, ERs, and jails. By the early 2000s, the Bush administration set a goal to end chronic homelessness within 10 years and funneled more HUD money into Permanent Supportive Housing (PSH) that followed Housing First principles. The Veterans Affairs department folded Housing First into its HUD-VASH voucher program. Congress cemented the shift with the 2009-2010 HEARTH Act, which merged funding streams, demanded performance metrics, and steered more dollars toward prevention, rapid rehousing, and permanent supportive units.

Housing First here will refer to two federal program types:

• Permanent Supportive Housing: A regular lease plus voluntary wrap-around services delivered by a mobile team. Tenant pays no more than 30% of their income for housing and the subsidy is indefinite.

• Rapid Rehousing: helps with deposits and a tapered rent subsidy (3-24 months) to get people out of shelter fast. Case management is lighter and time-limited.

Both start with a signed lease; treatment, sobriety, or employment are goals that are never demanded as a condition for tenancy.

Measuring homelessness

³Once a year, in the cold dark of late January (early February), armies of outreach workers and volunteers fan out across alleys, underpasses, and shelter dorms to take a one-night census of homelessness. Their clipboards feed the Point-in-Time (PiT) Count, HUD’s annual head-count of everyone sleeping outdoors and everyone bunking in a shelter, safe haven, or transitional program.

The PiT dates back to 2005, covers every HUD Continuum-of-Care (the counties or city coalitions charged with managing local homeless services), and is broken out by family status, veteran status, chronicity, and age range. Though imperfect, it is the closest thing the U.S. has to a national, year-over-year scoreboard.

That same night, each CoC files a Housing Inventory Count (HIC): a census of beds, not people. It tallies how many slots exist in emergency shelters, transitional housing, rapid-rehousing programs, and permanent supportive housing—and, crucially, the percentage of those beds that are filled.

Finally, HUD’s CoC Competition Award files show the money trail. Every project that wins federal funding is listed with its dollar amount and program type. Add those grants up and you get a yearly “budget pie” for each CoC: how much is going to Housing First programs (PSH or RRH) versus the old emergency or transitional slots.

There are some caveats:

• The PIT is a single winter snapshot that misses couch-surfers and people cycling through hospitals or jails, and different CoCs follow different counting playbooks (or the same CoC will change their playbooks over the years).

• Definitions of homelessness changed in 2013.

• Even with undercounts, these are still the only national datasets that roughly line up people, beds, and funding so anyone seeking evidence-based policy has to start here.

In sum, the PiT tells us the score, the HIC shows the playing field, and the Award files reveal how the coaches choose who gets to play and for how long. This triad lets us see whether the shift to Housing First is getting us to the “just right” policy mix or just stirring the porridge from the wrong bowl.

Is Housing First working?

A disclaimer before we look at some charts: nothing that follows supports a causal verdict. We won’t see whether homelessness would have been better or worse if HUD hadn’t shifted from Treatment First to Housing First. We’re simply walking through head-counts and bed-counts to what what’s been trending in the Housing First era. Treat it as a temperature check, not an experiment. That said, the picture that emerges is less flattering to HF than one may at first expect.

The overall count looks great, though more modest than it seems due to scaling. From the first fully comparable tally in 2007 through 2012, the headline number dipped 3-4 percent. After HUD’s 2013 definition change—and fresh funds for PSH—the decline sped up, falling another 4-5 percent through 2016. Victory laps were taken. Then the line bent the other way: between 2016 and the last pre-COVID count in 2019, homelessness ticked back up roughly 3 percent.

The standard Housing First narrative reads this chart as a win. Availability of, and access to, PSH and RRH units increases, and the total number of homelessness drops. Huzzah. Just look at the spending distribution by program over time.

Not only is spending steadily increasing; the share of spending on PSH, especially, jumps to around 80 percent of the total award (minus spending on HMIS and CoC planning grants).

But the dip in homeless is open to other stories. Unsheltered counts became more rigorous, rents outpaced wages in many metros, and local enforcement swings can push street visibility up or down without changing the underlying need. The national trend hints at success followed by a stall-out; it doesn’t say whether the stall is a supply bottleneck, a counting tweak, or a policy design flaw.

Advocates often cry that the supply of HF units is the problem. But, since 2007, HUD reports how many PSH and RRH beds are occupied on the night of the PIT count. Glancing at the graph below, you’ll see that the occupancy rate has hovered below 83 percent every single year through 2019. In words, roughly 1 bed in 6 is sitting empty.

This complicates the simple “we just need more units” narrative in two ways:

1. Slack. An 83 percent ceiling implies tens of thousands of beds are technically available. The bottleneck can’t be only a shortage of brick and mortar. At least some of the gap is people not showing up, not qualifying, or burning out of the program and leaving vacancies.

2. Goldilocks, again. If low-barrier programs still leave beds empty, either the gate isn’t as open as advertised (paperwork hurdles, landlord vetoes, geographic mismatch) or the inside of the program isn’t sticky enough (safety worries, treatment aversion, crime temptation) to keep residents from cycling out. In either case, pure capacity expansion misses the mark. Maybe outreach or program design needs reform.

No doubt, some vacancies reflect seasonal ebb-and-flow or lags in filling brand new units. But a persistent 17 percent hole in the occupancy rate (given persisting unsheltered counts) signals a deeper mismatch between what’s offered and what the unsheltered population will take. That’s Goldilocks’s bleeding heart.

To further isolate this concern, the next chart lays two lines on top of each other. Line 1 is the average heat-count of people sleeping outdoors. Line 2 is the average number of beds already filled in PSH housing. The picture is odd.

From 2012 to 2013, the street line dips a hair—almost certainly the blip from changing the terms. Starting in 2014, the PSH-occupancy keeps climbing at a steady clip, but the street line turns modestly north as well. By 2019 we have more people housed and more people unsheltered.

If Housing First is a one-for-one swap—add a PSH key, subtract a sidewalk sleeper—those two lines should move in opposite directions. They don’t. Here are some candidate reasons:

1. Some “newly housed” folks may have never shown up in the street count; they were couch surfing, living in cars, or teetering eviction and got pulled into PSH. If so, PSH may be keeping people off the street.

2. Outreach could be improving so the unsheltered census is catching people who were invisible in prior years.

3. Or the very low barriers of PSH fill units with people who aren’t the hardest-to-house so that the most vulnerable are still outside.

And, of course, more than one of these reasons may apply. But here’s a clean take-away: scaling HF by itself hasn’t generated a clean, visible drop in unsheltered homelessness. There seem to be unmet needs on the street and possibly mis-targeting.

But here’s my worry: Funding for subsidized housing is dwarfing (if not eliminating) funding for wrap-around services, including treatment. Most federal homeless-assistance dollars go to (1) renewing subsidies and (2) maintaining the data systems that are painfully hard to get access to. The day-to-day supports, such as mental health care, addiction counseling, employment programs, or even the street outreach budgets that get people off the streets are getting far less love. If wrap-around services are thin, even with weak incentives to participate, would we be surprised if tenants were drifting back on the sidewalks?

The graph below plots two lines. In bold are the total awards in millions. The dashed line cuts out HUD planning grants, HMIS data money, and the renewals for existing housing (PSH & RRH). What’s left are the dollars for new PSH/RRH beds, Transitioning Housing (new or renewal), and Supportive Services Only projects (again, new or renewal). The difference was about $300 million in 2007; by 2019 it had ballooned to nearly $2 billion—a sign that renewal costs are crowding out fresh service capacity. Street counts, recall, started climbing again around 2014. The timing isn’t proof of causality, but raises an eyebrow. As more cash is locked into renewals, fewer flexible dollars are free to fund the outreach and treatment supports that are models predict are critical.⁴

To conclude

People sleeping outside are rarely just short a rent check. Many wrestle with untreated mental health or substance abuse disorders, distrust institutions, or prefer the perceived autonomy of the street to a crowded, rule-bound (or otherwise risky) building. A purely housing supply lens assumes that every unsheltered person is queueing for a unit; the trends suggest a chunk of that population is balking for more complex reasons. Because they have little market power, their choices aren’t governed by rent levels the way standard housing demand is. Without credible outreach and on-site supports, new units risk sitting empty—or turning over rapidly—as tenants church in and out.

Is paying someone’s rent enough? Is there a place for a “housing readiness” condition? Unless the federal wallet funds safety, treatment carrots, and outreach that make a unit worth keeping, we may be stirring porridge in vein. The growing gaps in service funding suggests we’ve gone off-balance. The rising unsheltered line signals its time to adjust. After all, the counts of homelessness for 2024 are higher than any prior year since 2007. Clearly something needs to change.

1. One public policy group negotiated a data sharing agreement for nearly 4 years. That’s just the contract to get (limited) access to the data. IMO, this is a serious problem for transparency and accountability since many service providers receive your tax dollars. The idea that these providers are concerned with protecting their clients’ privacy is bogus since there are simply de-identification measures.

2. The federal government responded to the homelessness crisis of the 70’s and 80’s in 1987 with the Stewart B. McKinney Homeless Assistance Act of 1987, the first major legislation addressing homelessness in the U.S. The McKinney Act (later renamed the McKinney-Vento Act) created a range of targeted programs: emergency shelter grants, transitional housing programs, supportive services, and surplus food and property distribution for homeless assistance, among others. Over the late 1980s and 1990s, annual McKinney-Vento appropriations grew, and new program components were added and refined.

3. Thanks are due to Paul Webster for discussions on what PiT counts tell us.

4. The cynic in me worries about grift.