Saturday, March 07, 2015

Navel-gazing season comes early

Is it too early for bogus statisculating with election surveys? Apparently not:

WASHINGTON -- Hillary Clinton’s troubles are costing her politically, as potential Republican presidential rivals have inched closer to her in 2016 matchups, a new McClatchy-Marist poll found Friday.

Even if that development was interesting (and if you don't want to be derided for excessive reliance on the "horse race" frame next year, now would be a good time to shut up about it), you're going to have a hard time demonstrating that it's true. Which strongly suggests that it doesn't belong on the front page.

The former secretary of state fell below the crucial 50 percent level of support in one-on-one matchups against Jeb Bush, Marco Rubio and Scott Walker, and she was barely above that benchmark against Rand Paul, Rick Perry and Ted Cruz.

The technical journalistic term is "psychologically important barrier," meaning an arbitrary round number that impresses journalists but has no known relevance in real life. And unlike similar psychologically important barriers in the stock market, this one is an estimate: the "margin of error" carefully reported at the end of the story tells you how likely it is that a sample value of 49% (say, Clinton vs. Bush or Rubio in this survey) might represent a population value of 50% or higher. 

... Their competitive showings against Clinton “may tap into some concerns voters have about her,” said Lee Miringoff, director of the Marist Institute for Public Opinion in New York, which conducted the survey. “It gets us back to stuff people find unpleasant about the Clintons.”

Yes, they may -- which, even discounting the possibility that they also may not, doesn't directly address any current "troubles" that are costing her politically.

Clinton was first dogged by reports last month about millions of dollars her family’s foundation has received from donors based in foreign countries. Monday, the second day of the polling, The New York Times reported that Clinton had used a personal email server to conduct government business while secretary of state.

The email furor dominated political news all week. Wednesday, a committee in the House of Representatives that’s investigating the 2012 terrorist attack in Benghazi, Libya, subpoenaed relevant emails.

Think back to that first course in quantitative methods and the three things you need to start positing a causal relationship: The thing that was caused has to come after the thing that caused it, it has to have really happened, and any confounding explanations have to be ruled out. Thus, it's nice to note that the email tempest broke out while the poll was in the field; absent cheap and reliable time travel, it couldn't have cost anyone politically before it happened. The degree to which it "dominated political news all week" -- given the following day's intense navel-gazing over the Israeli prime minister's address to Congress -- is at the least debatable. And since we're comparing all the notional head-to-head matchups with similar estimates from December, we need to ask what else might have happened since then to influence potential changes in the level of support for Republican candidates -- say, the CPAC conference at the end of February.

Worse than that, quite a bit of polling since the previous McClatchy-Marist effort has found Clinton below, or below and above, the crucial 50 percent* pseudo-barrier. None of that polling is by definition better than McClatchy's, and in many cases (registered voters vs. likely voters vs. adults) it should be compared only with caution. But taken as a whole, it adds up to exactly what it seems: Lots of things might or might not affect survey results, especially a year before the horses are actually in a horse race. Any of them that actually happened here may or may not be some of them.

The results of the current poll aren't bad results (or dishonest ones, as in Fox's framing of the Iran issue). They're just not very interesting, and to the extent they're made sexy by card-stacking and other logical fallacies, they're misleading. If you want to do it right in 2016, kids, the time to stop doing it wrong is now

* The story does contain this line: "Experts consider a drop below 50 percent a danger sign for well-known candidates." Until some of those experts are brought in to explain and support the claim, you should ignore it.
d more here:



Post a Comment

<< Home