A three-step method for obtaining probability samples of homeless adults across diverse settings over large geographical areas was applied in two cities. Brief surveys determined that most homeless people (71-73%) had used a shelter in the prior year and a sizable additional group (18-20%) had used a food program, but not a shelter. Much smaller numbers were found at various in- and out-patient agencies. Less than 1% of the homeless surveyed on the streets used no services in the prior year.

These results support the researcher who focuses primarily on shelters and secondarily on food programs in order to obtain a representative sample. A profile of the homeless adult population based on samples receiving full-length interviews (N = 420 + 297 = 717) is provided and compared with prior studies using similarly rigorous probability sampling techniques. Relatively few and generally small differences between the samples from the two cities were found, suggesting that inconsistencies across studies are largely due to methodological rather than geographical factors. Similarly, few and small differences were found among participants obtained across different seasons and sampled from different types of sites (e.g., shelters vs. soup kitchens).