Gadflying Good Statistics

I’m an natural statistics skeptic. I disbelieve 78% of all statistics I read or hear.

Over at CafeHayek, Russell Roberts has some apparently good news for us, by way of the Grey Lady:

In 1972-73, for example, just 42 percent of the bottom fifth of American households owned a car; in 2003, almost three-quarters of “poverty households” had one. By 2001, only 6 percent of “poverty households” lived in “crowded” homes (more than one person per room) – down from 26 percent in 1970. By 2003, the fraction of poverty households with central air-conditioning (45 percent) was much higher than the 1980 level for the non-poor (29 percent).

Emphasis mine. On first reading, this looks like a tremendous development. But I bolded the words for a good reason: we see three different terms used in the same paragraph: “bottom fifth”, “poverty households”, and “non-poor”. My legal training tells me, when I see three different terms, there are three different things meant. There are exceptions, of course, but statutes, provision, and even judicial opinions are drawn with careful attention to language.

The question here is: are the “bottom fifth”, “poverty households”, and “non-poor” the same group of people? My spider sense tells me “no”. If they’re roughly the same, there isn’t much problem. If they’re not, the statistics are essentially meaningless.

EDIT: Thanks to the somewhat anonymous Anarchist in the comments, I should clarify that I meant to ask whether the “non-poor” be identified with the top four-fifths, the non-poverty households, or something else entirely.

3 thoughts on “Gadflying Good Statistics”

  1. Yeah, and of course he doesn’t say *where* he got these stats. He doesn’t seem to cite them in any of his relevant papers on his AEI page, either. Also, Google Print doesn’t find any matches for “eberstadt air conditioning” and Google Scholar doesn’t have anything for that search string that seems useful. I tried finding them on the Census Bureau site, but their navigation is a fucking disaster.

    I did manage to find percentage of households with income below the poverty level with more than 1.0 people/room by adding up various rows in the table “HCT22. TENURE BY POVERTY STATUS IN 1999 BY PLUMBING FACILITIES BY OCCUPANTS PER ROOM [39] – Universe: Occupied housing units” (like I said, their navigation is terrible)…unless my arithmetic is fucked, I get 13.3% of poor households being “crowded” using Eberstadt’s definition. Which is more than double Eberstadt’s number. Admittedly, he claims his figure is from 2001 and my number comes from the 2000 census, but it seems unlikely that a percentage like that would halve in a year. On the other hand, that’s still half of the claimed number for 1970 (the provenance of which I’m much too lazy to check up on).

    In other words, your skepticism seems to be well-founded.

  2. The question here is: are the “bottom fifth”, “overty households”, and “non-poor” the same group of people?

    Er… I don’t think the “non-poor” and the �poverty households” have much overlap by any definition. Heh.

  3. Reposted and hopefully corrected:

    The question here is: are the “bottom fifth”, “poverty households”, and “non-poor” the same group of people?

    Er… I don’t think the “non-poor” and the “poverty households” have much overlap by any definition. Heh.

Leave a Reply

Your email address will not be published. Required fields are marked *