• Choose a category
• All
• Classic-Fiction

By Root 761 0

for a number because we are Americans with western little minds that have to quantify everything in sight, whether we can or not.” But few reporters took note; instead, many repeated the “meaningless” 3 million estimate for years without conveying any sense of its spurious basis. More than seven years after Snyder confessed that his number was “meaningless,” for example, the CBS reporter John Roberts stated flatly that there were “more than three million homeless in America.”

LESSON: Not All “Studies” Are Equal

SOME QUESTIONS TO ASK WHEN THINKING ABOUT A DRAMATIC FACTUAL claim:

• Who stands behind the information?

• Does the source have an ax to grind?

• What method did the source use to obtain the information?

• How old are the data?

• What assumptions did those collecting the information make?

• How much guesswork was involved?

Today, we can say with some confidence that homeless people in the United States number in the hundreds of thousands, not in the millions. This is still a lot of people without a home, but it’s a fraction of what Snyder claimed.

In late March 2000, the U.S. Census Bureau counted 170,706 persons in homeless shelters and soup lines and at a number of open-air locations, such as under the Brooklyn Bridge, where homeless persons were known to gather. The Census Bureau says that figure shouldn’t be taken as a count of all the homeless—it concedes, for example, that it missed anyone who was not using a shelter on that early spring day, or who was sleeping in an open-air location other than those checked. But at least we can be certain that those 170,706 homeless persons were actually counted.

How many were missed is still open to question. Martha Burt, an expert at the Urban Institute who has studied homelessness for years, estimates that homeless people number no fewer than 444,000 and that the figure is probably closer to 842,000. Her estimates are projections based on an unprecedented, onetime Census Bureau survey of thousands of programs providing services to the homeless in seventy-six cities, suburbs, and rural areas. In October and November 1996, the bureau counted the persons being served at a random sample of those service locations and interviewed a random sampling of 4,207 clients to get additional information. The survey wasn’t intended or designed to produce a national estimate of the homeless population, and the Census Bureau didn’t attempt to derive one. But Burt made a few assumptions and calculated that nationally the number would have been 444,000 homeless adults and children using services on an average week in October and November, the months in which the Census Bureau conducted its head count. Burt also estimated that 842,000 homeless adults and children would have used services nationally on an average week in February, when the weather is much colder. Her estimates are just that—estimates—but they extrapolate from solid data, on the basis of assumptions that are fully disclosed.

Something else gives us confidence in Burt’s estimate that between 444,000 and 842,000 Americans are homeless: several other studies using different but still systematic methods have come up with figures that are generally in the same range. When different methods arrive at similar estimates, those estimates are more credible. We call this “convergent evidence.”

A Taxing Argument

As a general rule, the source of evidence matters. Snyder’s figures should have been viewed with a critical eye from the start both because he was lobbying for more federal spending to aid homeless people and because he was a bitter critic of the Reagan administration. While there are certainly scrupulously honest advocates out there, it’s clear that Snyder’s figures were “data in the service of ideology.” But we can place more trust in studies and data from sources that have no horse in the race, and who haven’t deceived us in the past. Such sources often give us a picture very different from the one painted by self-interested people who are trying to sell us something, whether it’s a product or a