Homelessness: Doing the Math

Hwy 87 Trail surveyTwo years ago, when the previous national Point in Time survey was done, Popular Mechanics magazine went back to the survey before that one in order to take a close look at how homeless veterans were counted. The first answer was that the Department of Housing and Urban Development had contracted with ABT Associates to count the people experiencing homelessness in America, including veterans.

ABT project director Alvaro Cortes explained how the numbers are massaged. This is called “imputing,” and it must be done because HUD knows how sketchy those reported numbers really are. For instance, volunteer population counters didn’t ask street people if they were veterans. And for some reason, HUD wasn’t counting any vets staying in VA shelters as homeless.

The official count is made every two years over a single 24-hour period. Some people are not found. Some don’t want to be enumerated. Joe Pappalardo describes the situation as it stood when he started researching the story:

In 2010, the actual head count of homeless veterans registered 61,011 people…. To reflect the homeless veterans the PIT count missed, HUD “imputed”– that is, they estimated — the number and added 15,318 homeless veterans to the official 2010 statistic.

In 2010 (and before), HUD’s imputations determined the number of unsheltered homeless vets by taking the percentage of the homeless vets reported in CoC shelters and applying that to the total number of unsheltered homeless tallied in the PIT. This added thousands of presumed homeless veterans to the statistic. The unadjusted 2010 PIT determined that 36,389 vets were in shelters and 24,728 were on the street; the adjusted 2010 count gives 43,437 sheltered and 32,892 unsheltered.

As of 2010, ABT had developed at least 10 different ways to “impute.” This imputing, Cortes told the writer, is to “reach the most accurate count.” Here is a basic question: How can they possibly determine which imputing algorithm produces the most accurate count, without knowing the actual number as a basis for comparison? And if they knew the actual number, there would be no need to impute. It all seems very arbitrary.

Start with the less confusing part, the numbers of street people reported by volunteers. It’s understandable why some adjustment might be needed. Take the number from the actual count and hypothesize possible scenarios that would augment that number. Apply an “imputing” formula and extrapolate a new number.

But why?

But. Why did the number of veterans in shelters need to be “imputed”? You’d think those numbers would be accurate. How many veterans are in temporary quarters under the jurisdiction of the VA? Isn’t the military famously meticulous about record-keeping? Were accurate records of shelter visits not being kept? And if not, why not? Those figures are of immense importance to the well-being of so many Americans.

Another question. Are patients in VA hospitals, with no place to be discharged to, counted among the homeless? Because they quite possibly might be, the moment they get out of there. Counting homeless veterans isn’t easy, and it seems at times that the underlying reason for the project is not “how can we get an accurate count?” but “how can we fiddle with the statistics to get a number we can convincingly brand as accurate?”

Pappalardo asked some of the same questions and pursued the answers. A lot of places do yearly counts of the homeless, and in 2011, volunteers were told to ask each interviewee if they were “a member or a former member of the U.S. armed services.” He wrote:

The way HUD counted veterans living in VA-run homeless shelters changed from 2010 to 2011 too. Veterans Affairs runs about 6000 emergency shelter beds in the nation; before 2011, CoCs were adjusting their counts to include these VA programs by taking the average number of beds that were occupied on any given night — about 86 percent in 2010 — and applying that across the board to get an estimate for the whole country. But those numbers didn’t match up to the VA’s statistics, and so in 2011, HUD instructed the CoCs and VA groups to reconcile the list and give a full inventory of beds for homeless veterans.

Okay, reconciling two sets of numbers is a good first step. Next: the part that critics had trouble with.


Source: “How Does Washington D.C. Count America’s Homeless Vets?” PopularMechanics.com, 01/19/12
Image by Richard Masoner