Originally published on January 22, 2022 in Los Angeles Daily News

By Clara Harter

Every year Los Angeles County faces the daunting task of counting its growing, ever on-the-move homeless population. While some view the count’s data as a precise science, others brand it a rough guesstimate at best.

The truth each year, despite the best efforts of a massive team of experts and volunteers, likely lies somewhere in between. And, while the methodology remains largely the same, this year the Los Angeles Homeless Services Authority is using a new counting application and hiring a demographer and two data scientists in an effort to improve the count’s accuracy following a missing census tract mishap in last year’s count.

The point-in-time count is coordinated by LAHSA and carried out by around 5,000 volunteers who travel the sprawling county by night over a three-day period. Their tally of tents, inhabited cars or RVs and people living outdoors is then combined with data on people living in shelters, a concurrent youth-specific count and three months of surveys by a USC team to paint a picture of Los Angeles’ unhoused population.

That data serves as a bellwether for how L.A. County is faring in its battle against homelessness, an application for county, state and federal funding and a map to help determine how resources should be deployed.

This year’s count will begin on Tuesday, Jan. 24 in the San Fernando and San Gabriel valleys, while east and west Los Angeles will be counted on Wednesday, Jan. 25, followed by South L.A., central L.A. and the Antelope Valley on Thursday, Jan. 26.

With great consequence and political capital riding on the results, the pressure is on to get it right.

That doesn’t always happen in practice.

Accusations of undercounting
When data from the 2022 Homeless Count was released last September, residents in the homeless hotspot of Venice were infuriated to see that the total number of tents, inhabited cars or RVs and people living outdoors in the Northwest corner of their neighborhood was zero.

This was no overnight miracle, but an error where no data was recorded for an entire census tract. The mistake both overshadowed and cast doubt on the exciting news that the rest of the Westside data captured – a 40% decrease in homelessness since the last count in 2020 that ran counter to the 4.1% increase seen countywide.

This error spurred L.A. City Council members to request an independent multi-year audit of the past homeless counts and a review of potential third-party vendors to complete the count in the future, in a November 2022 meeting.

“Obviously that tract has lots of people,” said Patricia St. Clair, statistical analysis director for the USC Homeless Count team. “I think we should have caught that.”The undercount stemmed from a series of errors, St. Clair said. The volunteer for that tract was meant to submit data throughout the counting process, but instead waited until the very end to do so at which point the counting application crashed, officials said. The volunteer then did not flag anything as having gone wrong when completing an exit survey and the USC and LAHSA teams failed to question the unusual data when reviewing results.

While the Venice example is an extreme one, the count has historically received accusations of undercounting and over time various groups such as Hollywood 4WRD, RAND Corporation and neighborhood groups have decided to run their own versions of the count in targeted areas.

New strategies aim for more accuracy
Part of the problem with the count’s methodology is that it was designed to meet a mandate by the U.S. Department of Housing and Urban Development to record accurate data on homelessness at the Continuum of Care (COC) level. (In the case of L.A’.s count, this means at the county level; Pasadena, Glendale and Long Beach, meanwhile, conduct their own independent counts.)

“I would say at that sort of level (the county level) probably what you’re looking at is one of the more rigorous counts in the country,” said Ben Henwood, USC Homeless Count team leader. “I think a lot of people rely on this for reasons other than estimating at the COC level — and that becomes a little trickier because it wasn’t necessarily designed to produce neighborhood level accurate estimates.”

LAHSA is well aware of the count’s past mistakes and the growing call for more accurate data at the local level, said Emily Vaughn Henry, deputy chief information officer for the agency.

To that end, there will be several tweaks to this year’s count.

First, LAHSA has hired its own of demographer and statisticians to review the count’s data.

“The way methodology was developed and calculated, it was never meant to inform at the census tract level,” said Vaughn Henry. “So the purpose of bringing a demographer and two data scientists is say ‘let’s see how we can start using that data to inform our council members so there’s confidence and there’s transparency in the data that we’re putting out there.”

In addition, this year’s count will use a new application developed by Esri, the trend-setting mapping and data company in Redlands, which has developed programs used in more than 50 homeless counts nationwide.

The Esri application is powered by GIS mapping technology and will be able to track exactly where volunteers walk during count night even if they lose cellular data.

“You’ll be able to see from our app the ‘bread crumbs’ all over the county, so you can get visual feedback that you’ve covered all of the intended areas,” said Este Geraghty, Esri’s chief medical officer.

This map of where volunteers walk will help LAHSA ensure every census tract is covered. In addition the application will provide updates to LAHSA every time a volunteer inputs data, Geraghty said, so that potential problems can be caught in real time.

Lastly, as a back-up measure and for volunteers without smartphones, traditional paper tracking forms will be given to everyone participating in this year’s count, Vaughn Henry said.

Counting what can’t be seen
Even with these new changes, the count still faces the challenge of capturing capturing data that is not visible to the immediate eye.

There are always worries that people hiding in alleys, dark spaces, bushes and other shelters will go unnoticed by volunteers. And, while the USC team of surveyors works to come up with accurate estimates on the average number of people living in tents, RVs and vehicles in each neighborhood, volunteers cannot physically see inside these structures.

“There’s definitely areas where there is a little less certainty,” said St. Clair. “For example, we know there’s not many unsheltered families and we really depend on our surveys to find them because they don’t tend to be on the street when our volunteers are out.”

Over the years the USC Homeless Count team has tried to better account for these ‘invisible’ people and for uncertainty by altering the error calculations in their analysis, St. Clair said.

“Overall, we do the best we can and I think, given the resources that we’ve had, the methodology is pretty solid,” she said. “There’s always little tweaks and there’s always some error, but overall, I think it gives us some idea of where we are.”

Measuring mental illness
Another aspect of the count that is hard to measure accurately is the number of homeless individuals experiencing mental health and substance-use issues.

In the 2022 Homeless Count results, LASHA stated that 39% of people experiencing homelessness reported experiencing serious mental illness or substance abuse — a data point based on survey results carried out by USC’s team in advance of the point-in-time count.

“I think both of these numbers are much bigger than what’s being reported,” said Supervisor Kathryn Barger in a statement released shortly after the 2022 results. “The California Policy Lab at UCLA, for example, found that the percentage of people experiencing mental health illness and substance abuse addiction is closer to 50%.”

Dr. Coley King, director of homeless health care at Venice Family Clinic, said that — based on his decades of experience providing street medicine to unhoused individuals — he would personally estimate higher percentages.

“I would say 15 up to 20% of the folks that are outdoors right now may suffer from a type of a serious mental illness – psychosis, seasonal affective disorder, schizophrenia, Bipolar disorder – and up to 70 or 80% have anxiety, depression and symptoms of trauma,” he said.

St. Clair, for her part, said she understood why some people thought the number should have been higher than 39%, but clarified that this measurement is based on the specific way the survey questions are asked.

In order to be included in that number people must answer affirmatively to experiencing mental illness and/or substance abuse and also have experienced this for a sustained period of time.

The questions are asked in this manner, St. Clair said, because people who say ‘yes’ to both queries meet HUD’s definition of being chronically homeless, which is a data point the county is required to report annually.

Henwood pointed out that Angelenos can sometimes overestimate the prevalence of mental illness among the unhoused because it can be shocking and memorable to observe.

“I think the perception is that there are high rates of mental health and addiction issues,” said Henwood. “On the other hand, that’s also what people notice and so there is an observer bias.”

Henwood also said that he recognizes many of the things that the count purports to measure can be politically inflammatory. Nevertheless, he said, the fact that the factors are mostly measured the same way, year-after-year, provides a useful reference point.

“I totally get the count can be a lightning rod, but it is an important project,” he said. “It’s one piece of data to help us understand homelessness, and it shouldn’t be the only piece of data. But it does allow us to see trends.”