Tuesday, May 20, 2008

How to lie with statistics

Over half a century ago, as a "senior elective" in high school, I took a class called
How to Lie with Statistics
, using a recently published book of the same title as our textbook. A remarkable class! And the book is still in print.

This rushes to mind as Clinton has become a textbook example of using these techniques in her misleading claim of winning the popular vote (which I've commented on before). Here is what she is doing:
  • She cites an accurate but meaningless statistic: at this point in time, there are more popular votes for her than for Obama. She is ahead by 26,967. (See
    RealClearPolitics
    ).
  • The statistic is meaningless because she counts Florida and especially Michigan, where O was not even on the ballot. However, in Michigan 238,168 voted for "Uncommitted" rather than for her. It is highly likely that over 27,000 of those would have voted for O had he been on the ballot, removing her lead.
  • She does not consider the popular votes of IA, NV, ME and WA, caucus states which did not report them. In other words, for all her rhetoric of including "everyone," i.e. FL and MI, she doesn't want to include the four caucus states won by Obama. If you project their popular votes based on delegate distribution, O again takes a commanding lead in "popular vote." Clinton uses the logic of Orwell's Animal Farm: "all states are equal but some are more equal than others."

In other words, the genius of this kind of blatant manipulation is that she's not "lying," strictly speaking, but everything she insinuates from citing a meaningless statistic does not follow from a neutral observation of all the data. This is how one lies with statistics. Politicians are doing it all the time.

I wonder if anyone is still teaching this class in high school?

Maybe in two weeks all this primary posturing will end and we can enter the arena of a new kind of posturing, the national election. Or will Clinton, still claiming the popular vote, take the fight to the convention? If she does, I think it ends her career -- so she probably won't. But we'll see. I wouldn't put anything past her, or past any politician.

No comments: