METHODS
UPDATED FOR 2026 Respondent Driven Sample
Every year as part of the Greater Los Angeles Homeless Count, youth homelessness is assessed through a survey-based enumeration from geographically sampled census tracts that are intended to represent the larger LA Continuum of Care. This method has relied primarily on youth homelessness service providers to physically canvas a large number of hotspot and sampled census tracts. It has become harder to implement due to capacity strain and competing priorities within the youth homelessness system. Further, some homeless youth service providers have expressed concern that this approach undercounts the homeless youth population because it does not cover all tracts, and it does not include people who are hidden or avoiding contact with the system. Given concerns about feasibility and undercounting, we will be replacing the previous geographically based sampling approach with a respondent-driven sampling (RDS) approach for YC2026 that will be both less resource-intensive and provide more coverage.
For this year's count, we will build on the published methodology for the King County Homeless Count and engage Zack Almquist of the University of Washington as a lead consultant for this work. For detailed background on the King County work, see: Innovating a community-driven enumeration and needs assessment of people experiencing homelessness: A network sampling approach for the HUD-mandated point-in-time count
What is Respondent-Driven Sampling?
Respondent Driven Sampling (RDS) is a network-based sampling method that uses social connections to reach and estimate hard-to-reach populations like youth experiencing homelessness. It has been used to reach populations like intravenous drug users, refugees, and other marginalized populations. Instead of trying to count a sample via geographic canvassing, RDS uses the social networks within the community to get a representative sample. We believe that a shift to this method will reduce the burden on youth providers and could potentially increase the accuracy of the homeless youth estimate.
RDS begins with initial interviews of respondents referred to as seeds at survey hubs. To ensure the broadest possible coverage, we will need to work from a large and representative number of hubs, though this will still involve far fewer locations than the tract-level canvas. After the interview, the seed respondents are provided with coupons that they are told to give to other people in their social circle that likely meet the criteria for the survey (i.e., unsheltered youth). Some respondents will have very few referrals, while some will have many. All coupons contain the ID of the person handing out the coupon as well as a unique ID for the recipient.
When a coupon recipient comes to a survey hub, they give the coupon they received to the interviewer. The information from the coupon is entered into a tracking system and the coupon recipient is interviewed. When the interview is over, the coupon recipient is then also given coupons to give to other people in their social circle to recruit the next wave in the recruitment chain. This process continues, recruiting further waves until the survey period ends or the chain is broken by recipients not returning their coupon.
Respondents are compensated for participation, and may also receive compensation for any coupons they distribute that are brought back to a survey hub by a recipient. The compensation is designed to encourage the distribution and return of the coupons so the coupon chain (and thus, the social network chain) can be built out. RDS relies on these chains to assess:
Referral ratios: The size of newly referred respondents compared to the initial sample.
Chain length: How long the referral chains are between respondents.
Reciprocity: The extent of mutual referrals between participants.
Multiple referrals: How often some respondents receive referrals from multiple sources.
This information is used to build out a visual map of the networks linking the entire sample, and to estimate how many people in the network were not sampled. This information can then be used to estimate the true size of the study population and their characteristics (e.g., race/ethnicity, gender, age).
Key Details of King County’s RDS Point-in-Time Count
A description of the application of RDS for the King County 2024 Point In Time can be found in their report: https://kcrha.org/wp-content/uploads/2025/05/Point-in-Time-Count-2024_King-County_final.pdf
Hub Locations
Nine hub locations were selected according to subregion based on previous PIT counts feedback from people with lived experience, conducting data collection from March 9th to April 6th, 2022
Extended time period (nearly a month) allowed for multiple waves of recruitment
671 interviews were collected over nine hub locations, with the largest chain being eight waves, resulting in an estimated 7,685 total unsheltered people experiencing homelessness
Multiple hub locations across different subregions made participation geographically feasible
Seeds
Seeds were obtained through the help of local service providers and appeared to be effectively random
These initial participants came to designated hub locations
Survey Completion
Participants completed both quantitative and qualitative surveys
King County used volunteers from the local Lived Experience Coalition (LEC), which describes themselves as a "largely BIPOC led" activist organization with strong ties to the community of people experiencing homelessness, to conduct all interviews
Upon completion, participants received $25
Recruitment Process
After completing their survey, each person was given three recruitment tickets/coupons. The ticket served as both:
Permission to participate in the survey
Transportation assistance (bus fare) to get to the hub location
Evidence of who recruited them (for tracking the network chains)
Each ticket also included bus fare to make it easier for recruited friends to get to hub locations
The participants were instructed to give these tickets to other people experiencing homelessness they knew
The typical recruitment size was three per a given contact's personal network
Chain Continuation
When the recruited friend completed their survey, they also received $25 and three new recruitment tickets
This process continued, creating recruitment "chains" through the social network