Sunday, August 14, 2016

Trust Political Pollimg

This is the quadrennial season when we get weedy about political polling.  This is also the season when, because it is political polling, some mug feels obliged to question the authenticity of the polls.  The polls are subject to the same defamation that infects most political discourse these days.  There are voices that expect a certain blind yassuh-boss gulping-whole of whatever drivel they’re selling.
Cynicism is easy.  Uncomfortable facts, we think, can be explained away.  Facts are not like that.  Facts are stubborn and ignored at our peril.
All we need is a little perspective.  Perspective in this case requires that we paid attention in math class and have a nodding acquaintance with the rules of logic.
That’s a tall order.  After all, ignorance is honored on many soap boxes and is a prerequisite to membership in some political movements.
That said, everything we need to learn is available to us stress-free.  Neither is it something we can’t learn in an afternoon.  Not that it comes cheap.  You don’t get to criticize if your basic knowledge is informed by half-assed opinion, wishful thinking or the utterances of those who aren’t clothed with expertise in the topic.
Terms like Margin of Sampling Error, Push Poll, Shy Tory Effect, Bradley Effect, Social Desirability Bias and Weighting will get you a long way.
Similarly, logical fallacies such as Post hoc ergo propter hoc and her kissin’ cousin cum hoc ergo propter hoc are useful principles as well.  They mean that just because events happen sequentially or at the same time doesn’t mean they can be linked. In the world of political slander these adages rank highest among the fallacies manifestly deserving of caution.

While we’re at it, correlation is not causation.  It’s the first thing one learns in logic class and the one thing least honored in the world of political gibberish.  Pundits have a way of filling-in-the-blanks between facts.  Two irrefutable facts thereby bookend a narrative of refutable humbug.

Take a look at this chart.  It appears to be all sciencey comparing the age of Miss Americas and homicide by hotness.  Maybe it is a silly example but you can easily see how false equivalence and mean spiritedness can conspire to produce an elegant lie.
When properly designed, polls do not lie.  Pollsters lie.  Poll respondents lie. Poll analysts lie.  Polls do not lie.  The adage attributed to Mark Twain is
particularly apt here: “Figures don’t lie liars figure”.
Here is an analysis of poll data that I call The Bowel College.   It purports to show one Electoral College victory path for Mrs. Clinton.  Mighty goofy but solid design by Nate Silver, It is included here because polls can be fun too.

So … armed with the healthy sort of skepticism and some basic knowledge, we can trust the polls.  With that trust, the news can be more about information and less about being hoodwinked.

No comments: