shelter in Centerville appeared especially compatible with the Hispanic female- headed households, who constituted the majority of the residents of the shelter. The women referred to the nuns as las monjitas (dear sisters) and generally had more patience with the rules of the shelter than did the non-Hispanic residents.
As a final strategy for the study, the research team reassembled for the March 20, 1990, Census Bureau Homeless Count (S-Night). Our goal was to count the homeless independently and then compare our findings with those of the Census Bureau. The comparison was done for people living outside, for shelter residents, and the residents of the hotel. We found that when the census takers visited the shelters and hotel in person, they received a high degree of cooperation from the management and tenants, and that the two counts were comparable. However, the census enumerators found no one on the street. It appeared that without prior relationships with the homeless living outside, there was very little likelihood of finding them. Since the doubled-up population was not to be covered by the S- Night count but by the regular April 1 census, it is difficult to know how many doubled-up people were actually counted.
Two levels of recommendations emerge from this study: recommendations at the federal level for improved methods of census enumeration for the homeless, and recommendations at the federal and local levels for improved strategies for moving people from homelessness to permanent housing.
Recommendation #1. Improve the street-count methodology. Nationwide, it appears that the homeless shelter count was reasonably accurate, due to a generally high degree of cooperation among shelter residents and managers. However, the street count fell short of most estimates. (See Wright 1992 for an excellent collection of articles evaluating the 1990 homeless count.) In Centerville, none of the homeless sleeping outside were counted.
The 1991 census in India contrasted with the U.S. census count of the homeless. The U.S. relies on sending out enumerators with no prior relationship with the persons being counted, in the dead of night. In contrast, India required that enumerators spend three weeks learning all the places the homeless ("houseless," in India) sleep within that enumerator's designated area of housed and non-housed residents. As in the United States, the homeless were counted at night, when they had bedded down on the pavement. The key difference was in the amount of preparation and knowledge of the homeless that the enumerator had. (See Glasser