How many people are experiencing unsheltered homelessness in King County? It’s currently an $11.8 billion question — the number floated by the King County Regional Homelessness Authority’s (KCRHA) draft five-year plan as what it will take to combat homelessness — and one that we haven’t always had an easy answer to.
The U.S. Department of Housing and Urban Development (HUD) mandates that local “continuum of care” agencies, like the KCRHA and All Home before it, do a count of unsheltered homeless individuals every other year. Up through 2020, King County performed a street count of people experiencing homelessness, picking a night in January and sending out hundreds of volunteers between 2 and 6 a.m. to count tents, bedrolls, buildings and RVs. Such counts are commonly referred to as point-in-time (PIT) counts.
The pandemic put the count on hiatus in 2021, but it resumed in 2022, albeit with a slightly altered methodology. This year, instead of doing the annual street count, KCRHA has elected to go with HUD’s every-other-year schedule.
“KCRHA weighed a variety of factors in making this decision: the unsheltered PIT count is generally considered the least accurate data collection by researchers and policymakers nationally; there is substantial and in-depth detail on estimations of the number and characteristics of people experiencing homelessness in the recently released Five-Year Plan; and the unsheltered count is not required by HUD this year,” said Anne Martens, the KCRHA’s director of communications, explaining the agency’s decision.
In characterizing the count as the “least accurate” way of measuring unsheltered homelessness, she is in good company. Most people working in homelessness advocacy consider it an undercount. By how much depends on whom you ask.
Alison Eisinger, the executive director of the Seattle/King County Coalition on Homelessness (SKCCH), which ran the count — formerly known as the One Night Count (ONC) — until 2017, estimated the actual number was sometimes as much as three times higher than ONC numbers.
“It's fair to say that the point-in-time count, as we organized it, was accurate but not complete,” she said. “And by virtue of its nature, a single point in time is never going to be the way in which you should be evaluating [these] things.”
What was it good for, then? How it connected the housed community with their unhoused neighbors, according to Eisinger.
“What I think this community needs, and currently does not have, is meaningful opportunities for ... people to have some deeper understanding of what it feels like to be outside without shelter in the middle of winter, in the middle of the night,” she said, “and to turn around and use that understanding to communicate with their lawmakers about the urgency of a real and sustained response.”
It also served as a helpful number for homelessness advocacy organizations looking to point to the urgency of the problem, often as part of fundraising efforts, she said.
“I can understand the reasoning by which the Regional Homelessness Authority has made the decision not to carry out an account of people without shelter in January 2023. And I don't disagree,” Eisinger said.
However, she also offered that her organization had elected to stop managing the count not because its methodology wasn’t sufficient, but because its admittedly incomplete numbers showed consistent double-digit increases in homelessness year-over-year, yet the scale of funding to fight homelessness never increased apace.
“We have enough data to tell us that our investments do not meet the scale of the need,” she warned.
KCRHA’s new model, referred to as the “respondent-driven sampling” (RDS) model, is more of a long exposure photo than a snapshot. It took place over a month and consisted of voluntary surveys completed by individuals experiencing unsheltered homelessness. Participants received a small cash incentive and were then given three tickets to give out to people in their encampments or social networks to also complete the survey. Any of those three who showed up and completed the survey were given three more tickets, which continued with each group until there were no more respondents. In the RDS model, the initial survey subjects are referred to as “seeds,” while each group of three they contact is a “wave.”
According to Zack Almquist, a University of Washington professor of both sociology and statistics who was part of the team that helped implement the RDS method for the 2022 PIT count, it was first developed for use by public health groups like the World Health Organization and the National Institutes of Health to study populations without what is called a “sampling frame” — something that makes them easy to delineate, like a home address or phone number.
If a map existed of every single encampment, RV cluster and rough sleeper in the county, it would be possible to do a census-style count, which is more akin to what ONC volunteers attempted in the past. However, given that such a map does not exist and that the unhoused population is amorphous and itinerant, it is much more accurate to shoot for a representative sample.
The people conducting the PIT worked with outreach workers to identify good seed candidates to begin the process. After planting those seeds, they waited for waves to arrive at nine different survey sites. There was attrition, Almquist said, but the project was ultimately successful. In one case, a single seed generated eight waves, he said.
To make sure respondents were comfortable participating, KCRHA staff and survey designers enlisted the help of the Lived Experience Coalition, a group of formerly homeless individuals who work with the KCRHA to guide policy. That, along with putting people experiencing homelessness in the driver’s seat of the process, really paid off, according to Almquist.
“One of the ideas with something like this one,” he said, is that, “we get a little more time to do it, which is nice. But the other, I think, really big thing is a lot of the population is hidden, and so you can actually persuade people to come in that would otherwise be really hard to find.”
All told, they got the representative sample they were looking for.
“It's like 650 interviews total, which is pretty big for an RDS. I feel pretty confident, and it worked pretty well,” he said.
Questions of quality aside, this also means that doing a classic PIT count in 2023 would likely yield a much lower number than in 2022. That could serve to further confuse the issue, as some politicians and policymakers already play fast and loose with PIT numbers.
King County Councilmember Reagan Dunn, for example, released a statement on Jan. 26, the day after KCRHA’s implementation board met to discuss the agency’s draft five-year plan, lambasting that plan for what he described as its “jaw-dropping” price tag. In it, he cited statistics from the 2019 PIT count to argue that “45% of the homeless come here from outside our area to take advantage of our generous services.”
The number appears to be a combination of two subsections of data from a survey conducted at the same time as the street count regarding length of time spent living in King County, broken down from within the 84 percent of all respondents who lost their housing while residing in King County. In 2019, 26 percent of respondents lost their housing after living in King County for one to four years, while 19 percent lost it after living here for less than a year. The survey did not collect information on why respondents moved to the area.
His second use of PIT statistics does make reference to the modernized 2022 count, which put the number of individuals experiencing homelessness at 13,368. Spending $11.8 billion to house that many people would constitute “a shocking $882,705 per homeless individual in King County,” Dunn’s statement read.
This is also less than accurate. As Almquist himself warned, even the 2022 RDS count is likely an undercount.
“I think anything we do will be on the low end of reality,” he said. “I think this produces a more accurate number and I think this is probably among the best you can do in the unsheltered population, but … I would treat it as a lower bound still and the traditional PIT count as an even lower estimate.”
Per the same document that lists the figure of 13,368 people experiencing homelessness in 2022, the number that KCRHA works off of is actually 40,871, the number of people who touched homelessness services in the past year. The figure comes from the King County Department of Community and Human Services’ 2020 “Cross Systems Homelessness Analysis.” That analysis, which combines data from the county’s Homeless Management Information System, its Behavioral Health System Data and our local Health Care for the Homeless Network, among others, combines these disparate datasets into one.
If the RDS count is a long exposure photo, the cross-systems analysis is a time lapse. And, if Dunn had used its findings to perform his “per person” cost analysis, he would arrive at a cost of $288,713 per individual over five years.
“That statement is a perfect example of why point-in-time count data alone are not sufficient,” said Eisginer, responding to Dunn’s comments on KCRHA’s five-year plan. “People fail to grasp the obvious but often overlooked element, which is [that] any single point in time is only that. It does not give you [data] over the course of a week, a month or a year.”
Martens declined to comment directly on Dunn’s statement, but said, “I will reiterate that [KCRHA CEO] Marc [Dones] has consistently said: ‘We have to talk about the challenges and the solutions at the right scale.’”
Figuring out what that scale is and crafting policy that provides sufficient resources toward it is, of course, the tricky part, but Almquist had some advice for policymakers: Aim high.
“If I had to fudge on things, I would say give more resources — because there could easily be some missing people — rather than give less resources,” he said.
Tobias Coughlin-Bogue is the associate editor at Real Change.
Read more of the Feb. 1-7, 2023 issue.