What is shelter diversion?

Diversion is a conversation at the front door of the homeless response system. A worker explores whether a household has any safe option other than shelter, and helps make that option work, sometimes with a small amount of money for a deposit, a bus ticket, a storage unit or a few weeks of rent. It is distinct from prevention, which reaches people before they lose housing, and distinct from shelter itself. In Maricopa County, applying a diversion strategy is a requirement for every coordinated entry access site.

Does shelter diversion work?

The honest answer is that nobody has shown it causally. The only randomized trial of diversion, run by the University of Notre Dame's Wilson Sheehan Lab for Economic Opportunities with Strategies to End Homelessness in Cincinnati, has published no results. The strongest evidence usually cited for diversion comes from a study of something adjacent: Chicago's Homelessness Prevention Call Center, which gave people money. Published diversion figures from programs themselves, such as a 49 percent diversion rate across 1,898 Puget Sound families, have no comparison group, so they cannot separate the effect of the intervention from the fact that most people offered diversion were never going to enter shelter.

What did the Chicago homelessness prevention study find?

Evans, Sullivan and Wallskog studied Chicago's Homelessness Prevention Call Center, which offers short-term financial help, usually under 1,000 dollars and typically one month of rent paid directly to a landlord. Shelter entry within three months fell by 1.4 percentage points against a baseline of 1.6 percent, an 88 percent reduction, and shelter days over six months fell by 2.6 against a baseline of 3.1 days, an 84 percent reduction. The design is quasi-experimental rather than randomized: it compares callers who reached the line when an agency had funds against callers who reached it when funds had run out.

Can diversion cause harm?

Yes, and the research that looked hardest for it found it. A three-year Canadian study of five youth shelter diversion programs, drawing on interviews with 30 staff and 47 young people, found that definitions of diversion varied between programs, that success was measured in non-comparable ways, that some young people were recorded as diverted only because the shelter was full, and that for some, diversion meant returning to an unsafe home. The researchers also had limited contact with youth who left for stable housing, so couch surfing, hidden homelessness and later shelter re-entry are poorly captured in the outcome data.

Evidence Brief

Maricopa Requires Diversion at Every Front Door. Its Rulebook Never Says What Diversion Is.

By Common Ladder · October 19, 2026 · 9 min read

Maricopa Regional CoC's Coordinated Entry Policies and Procedures, last updated by its board on June 11, 2026, require every Access Site to “Apply a diversion strategy.” The document never defines diversion, does not list it among assessment training topics, and never mentions financial assistance for it.¹ The CoC's 2026 federal application, by contrast, says it trains providers on diversion, runs diversion assistance and flex funding, and has released a regionwide Prevention and Diversion Toolkit — none of which the binding rule references.⁹
The strongest causal evidence anyone cites for diversion is not about diversion. It is about cash: Chicago's prevention call center cut three-month shelter entry by 1.4 points off a 1.6% baseline (88%) and six-month shelter days by 2.6 off 3.1 (84%) — by handing out roughly one month's rent, usually under $1,000, paid to a landlord.²
That study is quasi-experimental, not randomized. Its comparison group exists only because some callers rang on days the money had run out.²
Base rates break the arithmetic. Only 2% of families denied that assistance entered shelter within six months. In New York City, only 12.8% of families who applied for prevention services entered shelter within three years.³ Most people you divert were never going to enter shelter — which is exactly why diversion looks so effective.
The one randomized trial of diversion itself has no results. Notre Dame's LEO has been running an RCT with Strategies to End Homelessness in Cincinnati, randomizing callers to an offer or a four-week wait. No findings, no completion date.⁴
The headline program numbers are uncontrolled: across 1,898 Puget Sound families, 49% were diverted at $1,668 average, 82.6% had not returned within a year, median 37 days to housing. No comparison group.⁵
Diversion can do harm. A three-year Canadian study of five youth programs — 30 staff and 47 young people interviewed — found inconsistent definitions, non-comparable success measures, youth logged as diverted because the shelter was full, and some sent back to unsafe homes. Its 68% diversion rate applies only to youth reached before first shelter contact.⁶
In Maricopa there is often nowhere to divert to. The family waitlist passed 200 households this summer against a stated seven-week average wait; family shelter and transitional units fell from 769 in 2024 to 630 in 2025; the state cut $867,000. The 2026 count found 2,103 people in homeless families, the most since 2015. A 2023 evaluation found 94% of those assessed did not receive housing help.⁷

A mother calls 211. She and her two kids have been staying with her sister for three weeks and the sister's lease says four people, not six. There is no shelter bed tonight. The worker does not say no. The worker asks questions: is there anyone else, is there a way to make the sister's place work a little longer, what would it take.

That is diversion — the busiest intervention in the homeless response system, and the least examined. Everyone at the front door does it. Almost nobody can say what it is, and nobody has shown that it works.

A requirement the rulebook never defines

Maricopa Regional CoC publishes the rules that govern every access point in the county. The current Coordinated Entry Policies and Procedures — first adopted January 2018, most recently updated by the CoC Board on June 11, 2026 — mentions diversion exactly twice. Section 3.03 requires each Access Site to “Apply a diversion strategy” so that households able to use their own resources do not enter the homeless system. Section 5.02.01 says prevention prioritization should “divert those who have housing options to that option.”¹

That is the whole of it. The document never says what a diversion strategy consists of, what counts as a successful one, or how to record it. Section 5.01 lists the topics access-site staff must be trained on; diversion is not among them. No financial assistance is attached to diversion anywhere in it.¹

The pieces exist elsewhere. In its 2026 consolidated application to HUD, the CoC says it instructs providers on applying diversion, operates “eviction diversion, diversion assistance, and flex funding programs,” and has released a regionwide Prevention and Diversion Toolkit for cities, towns and tribal communities.⁹ But the application also offers no definition, and the binding rule names none of those resources. A worker reading the rule they are held to learns that diversion is required and nothing about what it is, or what they may offer.

The best evidence is about money, not conversation

When diversion is defended with research, the research is almost always Chicago's Homelessness Prevention Call Center. Evans, Sullivan and Wallskog published it in Science in 2016, and the effect sizes are genuinely large. Shelter entry within three months fell by 1.4 percentage points against a baseline of 1.6% — an 88% reduction. Shelter days over the following six months fell by 2.6 against a baseline of 3.1 — 84%.²

Two things matter more than the headline. First, the design: nobody was randomized. Whether a caller got help depended on whether an agency had funds left that day, and the authors use that unpredictability as their source of variation. A good natural experiment, not a trial.²

Second, and more important: the intervention was money. A typical household got one month of rent paid directly to a landlord, generally under $1,000, to clear a specific financial problem.² That is not a problem-solving conversation. It is a check. Citing this study to justify a diversion mandate that comes with no money attached inverts what it found.

Low base rates make everything look like it works

Here is the number that should discipline every diversion claim you read. Among the Chicago families who were denied prevention assistance, only 2% entered shelter within six months. In New York City, of all families who applied for prevention services, only 12.8% entered shelter within three years.³

Read that against a program reporting a 90% diversion rate. If nine in ten people walking through your door were not going to end up in shelter regardless, a program that does nothing at all posts a diversion rate in the high eighties. The measure is almost entirely base rate.

This is the central targeting problem in prevention work, and it cuts both ways: intervene early, while a crisis is cheap to solve, and you mostly reach people who would have been fine; intervene late, when risk is unmistakable, and the damage is done.³ Diversion sits at the earliest, least-targeted end of that trade.

The trial that would settle it has not reported

There is one randomized trial of diversion. Notre Dame's Wilson Sheehan Lab for Economic Opportunities is running it with Strategies to End Homelessness in Cincinnati: consenting callers are randomly assigned either to an offer of rental assistance and three to six months of case management, or to a control group not enrolled for at least four weeks.⁴

The project page names the researchers and the intended outcomes. It reports no results, no sample size and no completion date.⁴ Note what is being tested: again, rental assistance plus case management — money and staff time, not a conversation.

What the program numbers can and cannot show

The figures that circulate in diversion training decks come from program self-evaluation. The best-known set is from two Puget Sound pilots in King and Pierce Counties, published in 2018: 1,898 households with children, 49% housed quickly through diversion, $1,668 average cost, a median 37 days to housing, and 82.6% with no return to homelessness within a year.⁵

Useful operational numbers, not evidence of effect: there is no comparison group, so the 49% cannot be separated from the base rate above, and the 82.6% non-return figure covers only families the program housed and could still find a year later. And the $1,668 average means this version of diversion had a budget.

Where it goes wrong

The most careful look at diversion's failure modes comes from Canadian youth research: a three-year study across five shelter diversion programs, with a youth advisory committee, two universities, and interviews with 30 staff and 47 young people.⁶

It found that diversion meant different things at different agencies; that success was measured in non-comparable ways; that some youth were recorded as diverted only because the shelter was full that night; that some were diverted back into homes that were not safe, which the authors describe plainly as harm; and that several were diverted repeatedly. Crucially, the researchers lost touch with most youth who reached stable housing, so couch surfing, hidden homelessness and later shelter re-entry are largely missing from the outcome data.⁶

Their frequently quoted figure — that upwards of 68% of youth can be diverted — carries a condition that usually gets dropped: it applies to young people reached before they first contact an emergency shelter.⁶ The authors still call diversion promising. They also say the research needed to call it a best practice has not been done.

In Maricopa, diversion is running into an empty system

Diversion assumes it steers people away from a service they could otherwise have had. Here, that is now questionable.

ProPublica reported on October 9, 2026 that the UMOM-run family shelter waitlist passed 200 families over the summer, against an organization-stated average wait of about seven weeks. Family shelter and transitional units fell from 769 to 630 in a year as relief funds expired, while the 2026 count found 2,103 people in homeless families, the most since 2015. A 2023 evaluation of the county's coordinated entry system found that 94% of people assessed for housing help did not receive it, and that only 6% of assessed families got help with permanent housing.⁷

UMOM's chief executive described the practice in diversion's own language: “Let's try to figure your situation out first.”⁷ A 211 operator put the other side of it to a caller: “It's not me telling you no. It's the system.”⁷

With a bed and a deposit behind it, diversion is a genuine offer of a faster, less disruptive route. With a 200-family waitlist and no money, the same conversation is a queue-management tool wearing a best-practice label. The script does not change. What it is doing does.

What would actually settle this

Put the definition in the rule. Lift the toolkit's working definition into the policies, say what a successful diversion is and how it is recorded, and add diversion to the Section 5.01 training list. A mandate without a definition is unmeasurable by construction.

Connect the money to the door. Every study with a real effect size — Chicago, Cincinnati, Puget Sound — tested diversion with flexible funds. The CoC says flex funds exist; the access-site rule should say how much, who can authorize them, and for what.

Record the denominator. A diversion rate without a comparison group is uninterpretable. Programs that track what happens to households they could not serve — the design that made the Chicago study possible — would produce usable local evidence within a year.

None of this argues against diversion; a quick, dignified solution beats a shelter bed. The argument is narrower: diversion is carried by a requirement its own rule never defines, justified by a study of a different intervention, and measured by a statistic that mostly reflects who walked in. That is a thin foundation for the busiest decision point in the system.

Questions people ask

Is diversion the same thing as being turned away?

It is not supposed to be. Diversion should end with a safe place the household chose, and help making it work; being turned away ends with nothing. But the two are recorded identically in most systems, which is why the Canadian youth research could not always tell them apart.⁶

If I call 211 and get asked these questions, am I being denied shelter?

Not necessarily, and it is reasonable to ask directly. Two questions are worth asking out loud: am I on a waitlist, and if so where; and is there any financial help available to make the option we are discussing work. Marybeth Shinn, whose team produced the New York prevention research cited here, made the first point to ProPublica about Maricopa specifically: “The process should be transparent. People should know if they're on a waitlist.”⁷

Does Maricopa County have an eviction diversion program?

A pilot was announced on May 26, 2026, in partnership with the City of Phoenix, targeting Phoenix ZIP codes with the highest eviction filing counts, with development funded in the county's tentative FY 2027 budget. As of this writing the county has published no eligibility criteria, no start date, no program phone number and no application route, saying details “will be shared as they are available.”⁸ Treat it as announced, not open. For eviction help that exists today, start with free legal help in Phoenix and prevention and rental assistance.

Sources

Maricopa Regional Continuum of Care, Coordinated Entry System Policies and Procedures (first adopted January 22, 2018; most recent CoC Board update listed June 11, 2026). Section 3.03 (Access Sites) requires each site to “Apply a diversion strategy”; Section 5.02.01 (Homeless Prevention) refers to diverting “those who have housing options to that option.” We read the document in full for a definition of diversion, a diversion training requirement in the Section 5.01 assessment-training list, and any financial assistance tied to diversion. None of the three appears. The document states no effective date. Published policies change; this describes the version posted as of October 19, 2026.
William N. Evans, James X. Sullivan and Melanie Wallskog, “The impact of homelessness prevention programs on homelessness,” Science (2016), studying Chicago's Homelessness Prevention Call Center. Effect sizes as reported and re-stated in Evans, Phillips and Ruffini, “Reducing and Preventing Homelessness: A Review of the Evidence and Charting a Research Agenda,” NBER Working Paper 26232: shelter entry within three months fell 1.4 percentage points from a 1.6% baseline (88%); shelter days over six months fell 2.6 from a 3.1-day baseline (84%). The same review describes the design as quasi-experimental, identified by whether an agency had funds available when a caller rang, and describes the typical award as one month of rent paid to the landlord, with eligibility generally limited to problems solvable for no more than $1,000. Figures are quoted from the NBER review rather than the Science paper, which is paywalled; we did not obtain the journal article's page range and so do not cite one. Note also the mixed downstream findings the review reports from Palmer et al. (2019) on the same program: violent-crime arrests fell 0.86 points off a 3.7% control rate, while property-crime arrests rose 1.0 point off a 1.9% control rate three years out.
Both base-rate figures as reported in the NBER review cited at note 2: 2% of families denied Chicago prevention assistance entered shelter within six months (Evans et al. 2016); 12.8% of New York City families who applied for services entered shelter within three years (Shinn et al. 2013). The early-versus-late targeting trade-off is the review's own conclusion.
University of Notre Dame, Wilson Sheehan Lab for Economic Opportunities, “Shelter Diversion — Strategies to End Homelessness” project page (Cincinnati/Hamilton County, Ohio; researchers Mary Kate Batistich and Robert Collinson). Randomized design, treatment offered a program spot and control not enrolled for at least four weeks; eligibility under 15% of area median income with two or more places stayed in the past 30 days; intervention is rental assistance plus 3–6 months of case management. The page lists intended outcomes only — no results, sample size, publications or completion date, as of October 19, 2026. The $4,700 per-episode figure on that page is local background context, not a study finding.
Building Changes, “Diversion can help families exit homelessness quickly, simply and safely” (published April 16, 2018), reporting on King County and Pierce County, Washington pilots begun in 2014: 1,898 households with children combined, 49% housed quickly, $1,668 average cost per family housed, median 37 days to housing, 76% into unsubsidized rentals, 82.6% without a return to homelessness within one year, and one-third of housed Pierce County families receiving no financial assistance. The publication reports no comparison or control group and gives no per-county sample sizes or shelter-cost comparison; the “lower than other interventions” framing is not accompanied by figures for those interventions.
Canadian Observatory on Homelessness / Homeless Hub, “Shelter Diversion and Youth Homelessness: Does It Really Work?” (2025), reporting a three-year study across five Canadian youth shelter diversion programs with a youth advisory committee, two universities, diversion data from two youth-serving organizations, and interviews with 30 staff and 47 youth. Reported findings include limited peer-reviewed research on effectiveness, inconsistent definitions and implementation across programs, non-comparable success measures, youth diverted because shelters were full, harm from diversion to unsafe homes, repeat diversion, differing outcomes for Black, Indigenous, racialized, 2SLGBTQIA+ and disabled youth, and limited longitudinal contact with youth who reached stable housing. The “upwards of 68%” diversion figure is the authors' own data analysis and is conditioned on engaging youth before first emergency-shelter contact. The post also references a related study in MDPI Youth, but the sentence reporting that study's figure is garbled in the published post, so we do not quote a number from it.
ProPublica, “Maricopa County's Housing Hub Is a Dead End for Homeless Families,” (published October 9, 2026; we did not record a byline), based on interviews with 30 homeless families. Figures as reported there: waitlist above 200 families over summer 2026; UMOM-stated average wait about seven weeks, with some families reporting longer; family shelter and transitional housing units at 630 in 2025 against 769 in 2024 per federal data, with 139 units closed after COVID-19 relief funds expired; $867,000 in state funding cut that had helped roughly 40 families; 2,103 people in homeless families in the 2026 point-in-time count, the highest since 2015 and described as an undercount; a 2023 evaluation finding 94% of those assessed did not receive housing help and 6% of assessed families received permanent housing help. Quotations from UMOM CEO Monique Lopez, the 211 operator, and Marybeth Shinn (Vanderbilt University) are as published. The 2023 evaluation referenced is the Maricopa Regional CoC's annual coordinated entry evaluation; we cite ProPublica's characterization of it rather than the evaluation document, which is hosted on azmag.gov and could not be retrieved by our tools (the host disallows automated access).
Maricopa County, “Maricopa County Launches Eviction Diversion Program” news release (announced May 26, 2026). Partnership with the City of Phoenix targeting Phoenix ZIP codes with the highest eviction case counts; components listed as early identification, resource navigation and referral, exploration of mediation, and data collection; development funded in the county's tentative FY 2027 budget; context given as more than 80,000 eviction filings in Maricopa County Justice Courts for three consecutive years. The release states no eligibility criteria, start date, program phone number or application route, and says details “will be shared as they are available.” The only number on the page is the county main line, which is not program-specific. A separate county headline dated August 17, 2026 reads “Eviction Prevention Pilot Launches”; we could not confirm from the county's own pages whether intake is open, so this page does not tell anyone to call it.
Maricopa Regional Continuum of Care, FY 2026 CoC Consolidated Application (posted September 29, 2026). Section 1C-8 describes the CoC as instructing providers on “applying diversion within shelter settings,” operating “eviction diversion, diversion assistance, and flex funding programs,” working with philanthropic partners on “diversion and rapid resolution strategies,” and having “released a regionwide Prevention and Diversion Toolkit for cities, towns, and tribal communities” to “guide the development of diversion programs.” The application gives no formal definition of diversion and no funding amounts for these programs. The toolkit itself is hosted on azmag.gov, which our tools could not retrieve; we therefore do not characterize its contents.

This brief reports effect sizes as the source documents state them, including null and mixed results, and says in the footnotes where we could not reach a primary document — notably the paywalled Science paper at note 2 and the 2023 coordinated entry evaluation at note 7. No provider phone numbers or addresses are listed on this page by design; for help right now, call 211. Have a correction or an updated figure? Email corrections.commonladder@gmail.com.