New York City Annual Estimate of Homeless Is Challenged

The New York Times, May 22, 2007


A Columbia University professor hired by the city to help conduct an annual estimate of homeless people has disavowed the project, saying its methods cause the number to be understated.

The professor, Julien O. Teitler, said yesterday that he was withdrawing from the project because he did not think the city’s approach would arrive at an accurate figure.

Officials in charge of the annual estimate adamantly disagreed, saying that they had used the same methods every year since the project began in 2005 and that the methods were sound and rigorous.

In New York and other cities, counting the homeless has been difficult because the population is transient and elusive. The effort is also politically significant, as the numbers affect how much money is devoted to shelters and mental health and substance abuse programs.

New York City is three years into a five-year “action plan” announced by Mayor Michael R. Bloomberg to end chronic homelessness and reduce the street population by two-thirds, all by April 2009. The results so far are mixed. The number of homeless adults in city shelters has fallen noticeably since 2004, but the number of homeless families is at a record high.

So it came as welcome news on May 2 when the city announced that the third annual Homeless Outreach Population Estimate had shown a slight decrease in unsheltered homeless people.

According to the city, that number fell to 3,755 this year, from 3,843 in 2006 and 4,395 in 2005 — a cumulative decline of 15 percent.

The city’s Department of Homeless Services said the estimate “shows the city is on track” to meeting the goal of reducing unsheltered homelessness by two-thirds.

However, Dr. Teitler, an associate professor of social work and sociology, questions the reliability of those estimates, saying that city officials were “arbitrarily adjusting” figures “based on an assumption that turns a scientifically based method into a nonscientific one.”

Maryanne Schretzman, the deputy commissioner for policy and planning at the department, said she respected Dr. Teitler but rejected his interpretation of the data. “He has the right of his professional disagreement, but we also, as the city, have to keep with what our standards are in applying the same methodology year after year,” she said.

Both Dr. Teitler and Dr. Schretzman are skilled in quantitative methods; he holds a doctorate in sociology and she in social welfare.

The methodology used to arrive at the estimate is complex.

Officials divided each borough into small study areas. The study areas, as well as subway stations, were categorized as either high-density or low-density, based on whether unsheltered homeless individuals had been observed there.

Over four hours — this year on Jan. 30 — teams of volunteers were assigned to canvass all the high-density areas and stations, as well as a random sample of the low-density areas and stations.

In 2005, the city hired Kim Hopper, a Columbia medical anthropologist, to do a “quality assurance” program to bolster the accuracy of the estimate.

Under Dr. Hopper’s direction, Columbia recruited dozens of “decoys” to go to the same areas and stations as the volunteers. The decoys posed as homeless people.

The volunteers were instructed to ask people who were lingering on the street, in parks or in the subways or if they had a place to spend the night — unless the people were asleep, in which case they were not to be disturbed.

Decoys, if questioned by the volunteers, were instructed to identify themselves and to give the volunteers stickers to record their locations. Otherwise, the decoys were instructed merely to keep track of whether they saw the volunteers pass by.

At the core of the dispute is how to appropriately use the decoys in calculating how many homeless people are missed in each count.

Dr. Teitler, who took over the quality assurance program this year, says the city should look at what proportion of decoys were not counted. If, in a given area, 20 percent of decoys were not counted, then one should estimate that 20 percent of actual homeless people were missed, he says.

Under Dr. Teitler’s method, the estimate of unsheltered homeless people would have been 4,039, rather than 3,755.

Dr. Schretzman said it was more sensible to take the midpoint of two sets of figures — the proportion of decoys passed by altogether, and the proportion of decoys who were not questioned but who did observe volunteers walking by.

Dr. Hopper said the method was more defensible in the past than now because Dr. Teitler had recruited more decoys this year, assigned them to unambiguous locations, and ensured that they stayed at assigned locations.

Dr. Hopper said he preferred Dr. Teitler’s methodology, but he added, “This is a relatively new estimation technique — not in the mathematical sense, but in the sense of putting a procedure in the field.”

Copyright 2007 The New York Times Company