Studies of homeless individuals typically sample subjects from few types of sites or regions within a metropolitan area. This article focuses on the biases that can result from such a practice. We obtained a probability sample of 419 homeless youth from 41 sites (shelters, drop-in centers, and streets) in four regions of Los Angeles County (LAC). We found that restricting the frame to only certain types of sites or geographic regions biased the sample in terms of demographic and background characteristics as well as rates of risky behaviors. For example, homeless youth who can be found in shelters (vs. those who do not use shelters) and Hollywood (vs. other regions of LAC) tend to be younger, minorities, and engage less in risky behaviors. Recommendations are provided for evaluating the trade-offs between bias and cost in selecting a sampling strategy for studies of homeless individuals.