Not Much of a Retraction

Radley Balko writes

here’s a related statistic that’s pretty mind blowing in and of itself: According to the FBI, in 2011 there were 3991.1 arrests for every 100,000 people living in America. That means over the course of a single year, one in 25 Americans was arrested.

No, it doesn’t. This would only be true if we assumed that people do not get arrested more than once a year. The FBI Uniform Crime Reports (UCR) specifically caution against interpreting arrest data in the way that Balko does. Indeed, they make this warning so prominently within the UCR documents that it’s hard to miss.

Apparently, Balko was caught out by readers pretty quickly, because he updated his story later the same afternoon by adding this note at the very bottom:

CLARIFICATION: A few folks have noted that the 1 in 25 figure could be misleading, given that many people are likely to have been arrested more than once. Fair enough. It’s an average. The most accurate way to phrase it would be that in 2011, there were approximately four arrests for every 100 residents.

Yes, things that are not true “could be misleading”. But after acknowledging that his “1 in 25 Americans was arrested last year” claim was inaccurate, he left it as the screaming headline of his story.

Comments

  1. Warren Terra says

    It’s worse than that: his correction was literally to multiply his misleading statistic by one (by 4/4) – not to attempt to get the correct proportion of Americans who were arrested last year.

    Somehow I suspect the correct statistic would be sufficiently shocking to right-thinking people. But apparently I can’t trust him to respect his readers enough to care about finding it.

    • Freeman says

      It’s worse than that: his correction was literally to multiply his misleading statistic by one (by 4/4)…

      Not quite, Warren. Balko’s 1 in 25 number was also a linear reduction of the FBI’s “3991.1 arrests for every 100,000″ numbers, simply dividing both by 4,000. The mistake he made was going from “X arrests per Y Americans” to “that means X per Y Americans was arrested”. Due to multiple arrests of a statistically significant number of individuals, that inference is erroneous, which he acknowledged as soon as it was pointed out to him. His clarification: “there were approximately four arrests for every 100 residents”, corrects the mistake and states it in exactly the same terms as the FBI report, simply reducing from per 100,000 to per 100 while dropping the erroneous assertion that the number of arrests per 100 represents separate individuals arrested.

      … – not to attempt to get the correct proportion of Americans who were arrested last year. Somehow I suspect the correct statistic would be sufficiently shocking to right-thinking people. But apparently I can’t trust him to respect his readers enough to care about finding it.

      Of course it would. Multiple arrests aren’t likely to have skewed things drastically unless there are a lot of people getting arrested a lot of times every year. One out of fifty is shocking enough for me. But that’s beside the point — Balko was simply reporting an FBI crime statistic and made a common error in his attempt to simplify the numbers, which he acknowledged, corrected, and clarified. As far is I can tell, it isn’t even possible to extract information on the number of individuals arrested from the publicly-available FBI UCR data, it’s sure not there in the summary Balko reported and linked to, and the wording of the FBI disclaimer on the subject of multiple arrests of individuals strongly suggests to me that it is not. I googled for it and came up with nothing more enlightening than the FBI UCR. Apparently it’s not as easy as you make it sound. Are you saying we can’t trust you to respect your readers enough to care about whether or not what you accuse someone else of neglecting to do is even feasible?

      • Warren Terra says

        Oh, give me a break. Balko is paid to research and write about this stuff, which I’m not, and he’s presumably got all sorts of knowledgeable people on speed dial. No, I don’t have the statistic; it’s possible the FBI website doesn’t have it prominently displayed, or doesn’t have it at all. But Balko could ask around for the correct number (which the FBI must surely have); he’s actually paid to, and he’s and actual expert. Also, his correction was a crock; look at it again, if you will:

        A few folks have noted that the 1 in 25 figure could be misleading, given that many people are likely to have been arrested more than once. Fair enough. It’s an average. The most accurate way to phrase it would be that in 2011, there were approximately four arrests for every 100 residents.

        As a correction, it’s OK in the middle, and then he slips, and makes almost exactly the same mistake at the end that he originally made (though instead of saying 4 per 100 people arrested he says 4 arrests per 100 people). If he’d left off the last phrase, it might have worked; as it is, he’s just confusing his readers. I can’t even figure out why he’s doing it.

        • Keith Humphreys says

          I agree with you Warren that this is a serious policy issue and simply telling the truth would have been sufficient to generate a needed policy discussion. People who feel that shading the truth or outright lying is justified if it supports their political agenda underestimate how much people they might otherwise persuade are repelled by dishonesty.

          • Freeman says

            If “approximately four arrests for every 100 residents” is “dishonest”, how is it that the FBI’s statistic “3991.1 arrests for every 100,000 people living in America” is not?

          • MobiusKlein says

            It’s dishonest because it does not answer the headline premise – x% of residents were arrested in 2013. If Balko quoted the actual numbers, it would be proper correction.

            As it is, the correction is half way there.

      • marcel says

        Multiple arrests aren’t likely to have skewed things drastically unless there are a lot of people getting arrested a lot of times every year.

        Nope. If there are a few people being arrested a gazillion or 50 quintillion times/year, then the distribution is even more skewed than in the example above, and the average number of arrests/100K people each year will be even further from the average number of people arrested/100K people each year.

        It is because of confusions of this sort that it is important to get this correct. What is needed is not the total number of arrests/year in the U.S. but the total number of people arrested at least once/year. Unless nobody is arrested more than once, the 2 means (over the total population) are not the same.

  2. alkali says

    This seems unnecessary. He made the retraction. Fiddling with headlines after the fact is tricky with some blogging software.

    • politicalfootball says

      If he didn’t correct the headline, he didn’t correct the headline. It strikes me as surpassingly unlikely that it’s impossible to correct a headline, but if so, he should say so and apologize for his failure to make the correction – not pretend that he’s done it.

      • alkali says

        But he did make the correction! That he hasn’t also corrected the headline to conform to the correction is regrettable, but does it matter so much?

  3. Mitch Guthman says

    Even with the “retraction/correction,” I have to agree with Warren. I don’t any effort to arrive at a correct figure of the percentage of Americans who were arrested last year. Also, just intuitively, I don’t think the FBI figure represents an “average,” either.

  4. Business ethics says

    But after acknowledging that his “1 in 25 Americans was arrested last year” claim was inaccurate, he left it as the screaming headline of his story.

    Cut him some slack.
    He is just another American citizen trying to sell himself in the marketplace.
    Nothing abnormal here…

  5. Michael O'Hare says

    This mistake pops up regularly in other contexts: museums with X visits per year often boast that “X people per year visited the museum in 20xx ” even when X is comparable to their metropolitan area adult population.

    • Warren Terra says

      One similar instance I find irritating: there’s this service called “researchgate” that tries to be a sort of literature-centered social network/profile-building/literature-sharing site for the scientific community (well, at least the biological research community, I don’t know how far they spread). They seem to be getting some uptake, and I validated my name and bibliography one idle evening, after I’d found their site to be a convenient place to get a somewhat obscure paper. So, now they email me about three times a week (thanks, guys), and every week one of the emails says “(your name), (large number) of your publications were cited”. Which is exciting, especially as I’ve published a fraction of “large number” times – what they mean is, my papers have been cited (large number) of times. As in, ever, so it’s not really news to me, especially when that number doesn’t change from week to week.

      • Anonymous37 says

        The whole “-gate” suffix has been completely screwed by unimaginative headline writers who needed some pithy way to say SCANDAL-associated-with-this-item.

        Which doesn’t really say anything about the site you mention, except that 1) good for them for refusing to kowtow to this moronic usage and 2) the association is so Pavlovian with me that I immediately scanned your comment for the bit where a scandal occurred.

    • James Wimberley says

      The obvious and quite cheap fix for museums is to survey a sample to get a handle on the repeat visitors. They should do this anyway as proper market research on their customers, not just for the headline. How many do? I’ve never been buttonholed by a survey-taker in a museum.

  6. ferd says

    Agree with Mitch Guthman above. ” ’3991.1 arrests for every 100,000′ in 2011″ is not “an average,” is it? It’s just a ratio for the year 2011, I think. Collect two of these ratios from two different years, add them together, then divide by 2, and THAT’s an average . . . I think.

    What’s it called in math when you modify a fraction so that the denominator is nice and round? ‘Simplification’? ‘Factorization’?

  7. capnhook says

    ^^^
    “What’s it called in math when you modify a fraction so that the denominator is nice and round? ‘Simplification’? ‘Factorization’?”

    It is called “the wrong answer”…..