Counting the Homeless

A modified capture-recapture methodology was devised to monitor and assess the 1990 S-Night street count in lower Manhattan. Observers were deployed at 41 sites, selected in a quasi-random manner, from the list provided of "predesignated" street sites. Such "plants" were situated similarly to the street-dwelling homeless and should themselves have been counted Results suggest that Census Bureau enumeration of homeless people in such sites fell short of a full count. Problems included the difficulty of counting sleeping people hidden from view; the incomplete listing of sites where homeless people congregate; faulty coverage at the sites visited by Bureau enumerators; the apparent failure of enumerators to visit many of the smaller sites; and the evident reluctance of some enumerators to follow Census Bureau procedures. Attitudes of the homeless poor themselves were not a problem. Implications are drawn for alternative methods of enumeration.