Data collection is difficult. Accurately counting homeless people.

To count the unhoused and unsheltered population—the shelters are usually full to bursting with waitlists hundreds or a thousand names long—county health or human services agencies, or nonprofits to which the task is contracted out, often resort to the simplest method of enumeration known, the one you learn in kindergarten: They (or citizen volunteers, mostly) go out with flashlights, clipboards and pencils, and literally count heads, or curled-up street sleepers, or RVs, or tents.

How many people an office manager or sales rep guesses are sleeping in an RV or a tent they’re peering at in the semi-dark becomes data. Whether the volunteer presumes two or four is up to them—I can tell you this, for I have done it twice, in 2009 and 2017, and I don’t believe my guessing skills improved much—and thus wholly arbitrary, a snap decision that can result in a variance of 100 percent. Or more. Is that just some old car, or an old car someone lives in? Is that RV the glamping vehicle for an Instagram influencer or some eccentric Burner type, or does it house the family of four who couldn’t afford the landlord’s latest offer? You don’t know and you can’t know. Yet, this is the data the federal government uses, and we arrive at neat numbers like, “500,000 homeless people in America, 8,011 homeless people in San Francisco.”

How California’s Homeless Crisis Grew Obscenely Out of Control, Chris Roberts,