22 August 2007

They say nobody reads anymore

John Armstrong is begging me in a comment at at Concurring Opinions to comment on this article from the Associated Press, which makes the following claims:

  • One in four adults say they read no books at all in the past year

  • "The typical person claimed to have read four books in the last year -- half read more and half read fewer." -- so although they don't want to use the word "median", they're saying the median number of books read is four.

  • "Excluding those who hadn't read any, the usual number read was seven." Assuming that "usual number" means "median", this is saying that five-eights of people (the quarter who read no books, plus half of the rest) read seven or less books in the last year.

So what do we know about the distribution? One-quarter of people read no books; one-quarter read between one and four; one-eighth read between four and seven; three-eighths read more.

They claim a 3% margin of error, as well, which is standard for polls involving a thousand people (as this one was), but that margin of error only applies to the survey as a whole. The article includes a lot of claims of the form "Xs read more than Ys", but the number of Xs or Ys that were polled is less than a thousand, so the margin of error is greater.

It seems hard to get these numbers, though. The article claims that "In 2004, a National Endowment for the Arts report titled "Reading at Risk" found only 57 percent of American adults had read a book in 2002, a four percentage point drop in a decade. The study faulted television, movies and the Internet." (Emphasis mine.) If you interpret both of these claims at face value, 18 percent of non-readers have been converted to readers in the last five years. This seems unlikely.

I keep a list of the books I read; there are approximately one hundred and seventeen books on it. Yes, you read that right. That number's a bit inflated by the fact that about a third of those books were books I was rereading. But it's also a bit deflated because I have a tendency not to count textbooks, research monographs, and so on. (There's a pile of books up to my knee -- mostly library books, which is why they're in a pile, so I remember to return them -- which I read in the last year but aren't on this list.) I've also probably read a dozen or so books while sitting at the bookstore because I was too cheap to buy them, and some book-length online works...

Of course, I'm not typical.

A "book" doesn't seem like the right unit here, though. For one thing, some books are much longer than others. To take the two books on my shelf that I suspect have the most and fewest words, I estimate that Victor Hugo's Les Misérables has about 700,000 words, and Susanna Kaysen's Girl, Interrupted has about 40,000. For another thing, the person who reads a lot of books is going to engage a lot more deeply with some of them than others. There have been books that I've read in two hours and never gone back to; there have been books that I've returned to over and over again and find new insights every time. Most, of course, fall somewhere in between. And what if you don't finish a book? Does it still count?

I think the more useful metric would be "how much time have you spent reading books in the past year?" Because you can't ask people that, I suspect the Right Thing to do is to call people up and say "how much time have you spent reading books in the last week?", doing so at various times of year to control for the fact that reading is probably higher at some times of year than others, and averaging the results. But no one cares that much. (Actually, I suspect people in the publishing industry care very much, but they're not releasing their findings if they have any.) Or perhaps "how many words have you read in the past year?", but counting this seems almost impossible. (It wouldn't surprise me to learn that this number is rising, as a lot of online content is in written form.)

And as "Katie" commenting at Concurring Opinions points out, anyone at the high end of the spectrum probably underestimates. I don't know if this is true. I would have estimated I read "about two books a week" in the last year, for 104 books in the year; when I went and looked at the list, the actual count was 117. This might not matter all that much, because the pollster was asking about the median, and the people who are going to have trouble are the ones that are above the median. But I suspect that this poll is a lot like the recent polls about the number of sexual partners; people feel that they should lie about the number of books they read. I'm not sure in which direction they're likely to lie, though; I suspect it's correlated with education, and also with how many books one thinks one's friends read.

And what's so great about books, anyway? Why do we assume that reading books is automatically better than reading any other source of the written word? I suppose the argument is that a 100,000-word book requires more intellectual effort to read than, say, one hundred 1,000-word newspaper or magazine articles, because there is more interrelation among the ideas. But books come, for the most part, predigested. A lot of the real intellectual work is done in taking those clippings from various sources and making a book out of them. But this isn't a study of how intellectuals read, it's a study of how the person in the street reads. And even "study" is a bit too strong. They called up a thousand people and asked them some desultory questions. The AP did the poll itself. Let's face it, they're just trying to sell newspapers.


Anonymous said...

What about audio books? About a year or two ago I tried several to see if you could really listen to a book; I concluded you could. Though I still like reading paper; I like the option of writing in the margins and underlining [obviously I don't collect rare books].

As newspapers are finding out, reading is not just an activity that revolves around paper and ink.

A list of books read!! Someone more anal than me!
More than lists I love to throw things away except books and records/CDs; my wife is a collector of things.

There was another survey about reading in the last 3 years which also said no one was reading - however it turned out it was about fiction; they did not brother to survey non-fiction.

An aside: The U.S. is the magazine country. We read more magazine than other countries especially European. Women read more magazines than men. Europeans read more newspapers than people in the U.S. These are all things I read several years ago, though I bet they still stand.

Is reading September Vogue equal to reading "In Our Time"? Is reading the Bell Sys Technical Journal on Unix and C equal to reading War and Peace in English?

meep said...

Also, you can't do the 3% margin of error when you've got more than yes/no going on - the 3% margin is for a binomial situation, not a multinomial situation. There's (negative) covariances between the items.

Just a thought on the next time poll results, when there's more than two choices, are quoted.