Wednesday, March 24, 2010

How to lie with statistics

A triviality, for today....

In my days at YU, I took a Sociology class for which we read a book called How to Lie with Statistics. The book was a fun read, but it had nothing on a column I saw yesterday, Communicating in 2010: Why leaders cannot ignore the impact of social media, which used (without attribution!) statistics from Socialnomics.

The article consists largely of a set of statistics selected to show that social media outlets in the age of the Internet (a) have grown quickly and (b) now reach a lot of people. Neither is a particularly novel or controversial point... which, perhaps, is what drove the author to use inappropriate comparisons and conclusions in order to say something that would grab people's attention. It's like a rabbi in a speech, extending himself to greater and more outlandish hyperbole in an attempt to grab people's attention.

Here are examples:

* It took radio 38 years to reach 50 million users, television 13 years, the Internet four years, and the iPod three years. In just a nine month period, Facebook added 100 million users, and downloads of iPhone applications reached one billion.

- The Facebook total is cumulative, including people who joined and never used it, or who joined and left and re-joined with another name. The radio and television users are simultaneous - 38 years after radio's inception, 50 million people were listening;

- iPhone applications are not one-per-person, and therefore they don't reflect an accurate means of comparison with the other items.

* Print newspaper circulation is down 7 million over the last 25 years. But in the last five years, unique readers of online newspapers have increased 30 million.

- Print circulation is down 7 million from a very, very large number. (Anyone know it?) Online newspapers are up from 0.

* Collectively, ABC, NBC, and CBS get 10 million unique visitors every month, and these businesses have been around for a combined 200 years. YouTube, Facebook, and MySpace get 250 million unique visitors each month, and they’ve only been around for the last six years.

- The unique visitors to networks are people who watch lots of television, at least 30-minutes worth. The unique visitors to MySpace include people who look at just one page.

- 'Unique visitors' to networks are gauged by household, and so 10 million may actually be 40 million or more. 'Unique visitors' to Internet sites are gauged by individual.

- I am not at all clear on why we are combining the network ages at all, but if we are then we should also combine the ages of Youtube, etc. for consistency.

* In 2008, John McCain raised $11 million for his U.S. presidential bid through traditional campaign fundraisers. Barack Obama leveraged online social networks to raise $55 million.

- Note: This was not on the Socialnomics site; don't know where she got this one.

-Why are we comparing apples (McCain and traditional fundraisers) to oranges (Obama and online fundraising)? Probably because Obama raised some $650 million dollars overall, which makes $55 million via social media look a little lame.

- Also unclear: McCain raised more than $30 million for his campaign - so did 2/3 of his fundraising come from online social networks? I'm confused.

* More than 1.5 million pieces of content are shared on Facebook daily.

- Define "content"!

* 80 per cent of companies are using LinkedIn as a recruitment tool.

- Going back to the data source shows they are actually using the site as a way to do research on prospective hires, once they have located the candidates in more traditional ways.

And so on.

Who cares? I do; I like statistics, but I also like honest presentations. Seems I have a תרתי דסתרי problem here. [תרתי דסתרי doesn't really translate into English; best I can do is "inherent contradiction.]


  1. The book you mentioned is a good one but you might also enjoy reading the books by Joel Best--first one was "Lies, Damned Lies and Statistics."

    Not the numbers' fault, but statistics can be used to show anything you want them to show depending on how you manipulate them. One of my peeves is those who say that the statistics show that a vast majority of people believe X or feel X but don't show you the numbers. Does 52% of the sampling (the size of which they never give you) really constitute a vast majority? Given an error margin of a few points and what we have is something other than what we are being told.

    The members of Klal also tend to use these statistics in a strange manner. We apply crisis to anything that is a problem despite not having accurate figures to back up such a designation.

  2. Ah, yes. Mareleyn Schneider's Intro to Sociology. Good class, not particularly quantitative, though. It's about 20 years since I took it; still have the book.

  3. tartei d'satrei-I always use internally inconsistent

    ProfK-we also don't seem to want to get statistics but rather prefer anecdotal data

    RH-BTW sounds like you might want to consider an actuarial career (those guys can make numbers sing any tune)
    Joel Rich

  4. Very good post. I'm afraid I don't understand why you consider תרתי דסתרי really untranslatable though. It's not like "ukimta" or "kugel," at least to my mind.

  5. s, how about inyan as in "there's an inyan to"
    Joel Rich

  6. s, how about inyan as in "there's an inyan to"
    Joel Rich

  7. ProfK-
    Ah, yes. Not to mention the definition of the words to which they agreed. As in Ha'Aretz's finding that some 40% of Israelis felt President Obama was fair to Israel, when they actually said he was עניני.

    Yes, that's the one.

    I suppose that's a fair translation, yes.

    And let's not forget "over," as in "he was over a lav." Which sounds like something a Brit would say regarding a friend with a hangover?