Thursday, April 3, 2014

Do We Vote To See If the Polls Are Right?

To those who impart great import into polls and the needs of policies that support the public good, observe an example from the findings of a Gallup Poll from 1954 (click to embiggen):


My father, a mathematical genius in his own right, would occasionally comment upon some poll result or statistical analysis that would stumble across our path as we went about our daily lives, usually from the newspaper or on the television after we acquired one.  He would deftly dissect the purported or advertised conclusion in a sort of succinct soliloquy to demonstrate that it was in error or (more likely) catered to a prior agenda.  It was from him that I learned the term a priori (existing in the mind prior to and independent from observation), the common fallacy of looking for and stacking evidence in such a way that it supports a preconceived conclusion.  More than once I would hear him mutter the observation of Mark Twain (who quoted Benjamin Disraeli): "Lies, damned lies, and statistics."

As I moved about my interrupted academic forays during my extended lifetime, I ended up taking political statistics courses three different times, because I was assured that each was Completely Different™ (not true) and a required course.  I have met more people than I care to remember who are exclusively concerned with 'framing' a problem and who become peckish when reminded that actual reality has a far different opinion on the matter.  Apparently, many of them have moved over into the cash cow of 'climate change studies'.

The realm of pop science is riddled with such fallacies, sustained through declarations that the "science is settled!"  And all by credentialed experts, no less.  That falls into such realms as the settled science of anthropogenic global warming (which would include the crisis of global cooling in the 1970s), Bohr's atomic model, the Steady State theory, the eugenics policies of Margaret Sanger and Heinrich Himmler, physiognomy and phrenology, repressed memory and dissociative identity (multiple personality) disorder, and so on.

So how much more flawed can polls be?  How many push polls are reported as factual findings?  How many polls do you see (assuming that you read the fine print) cite a total number of respondents at less than 500 but nevertheless have a margin of error of ± 3% (stick your tongue firmly in your cheek on those)?  How were those respondents queried ('man in the street', voluntary response or open access polls [e.g., online polls], door to door, random access)?  Is the question on your TV screen actually question number 72 on the survey, and how did the previous 71 questions set the tone that helped shape the answer?

The standard method is through random phone polling (once we got past the infamous "Dewey Defeats Truman" headline of the Chicago Tribune).  Computer algorithms are created to ensure randomness of telephone calls, and this is still considered the best method for polling, but the public at large is not generally aware that the results are often further massaged to fit a pre-set model of what the pollsters consider to be a proper distribution of variables, such as party affiliation.

And how well is the phone poll system keeping up with changing technology?  How many people answer their home phones and proceed to answer questions?  What sort of people are available to do that – housewives, unemployed, shut-ins, political junkies? How many people have opted out of home phones altogether in favor of cell phones?  How many people screen their calls to avoid such nuisances?  How many calls go automatically to pagers (there are still some)?  Can other algorithms keep up with an increasing ability for people to revert to an electronic version of a 'man in the street' interview?  In other words, how many people have the poll takers missed (and what is their distribution) who have opinions that lead them to vote?  How many remaining people who answer the poll questions do so because they are politically motivated to do so?

Don't get me wrong – there are some polls that are better than others, sometimes far better.  But we have been schooled into accepting that these polls have an accuracy that is, shall we say, optimistic.  How many of them result in an increasingly accurate measurement of an estimate?

So, do these polls measure a valid determination of what the public feels is an important cultural touchstone, or what the mavens of pop media has driven the public to respond to?  We often hear the snickers of the pundits who report on the percentage of people who believe in UFOs, or who believe that Elvis is still alive, or who are convinced that man walking on the moon was a hoax.  How many perverse answers are built into responses to questions such as these?  ("This clown is asking me if I believe the Earth is flat?  WTF?  Okay, I'll answer 'yes' – that'll screw them up.")

Yet after we shake our head at the distressingly high percentage of strange beliefs, we turn the page and regard the poll about how many people demand that the 'rich' should pay an increasingly higher 'fair share' of the public debt, or who believe that water running in streams should somehow be pristine in a state of nature (and are willing to spend scads of other peoples' money to make it so).

Yet these people vote.

No comments:

Post a Comment

Comments are welcome and discussion is open and encouraged. I expect that there will be some occasional disagreement (heaven knows why) or welcome clarification and embellishment, and such are freely solicited.

Consider that all such comments are in the public domain and are expected to be polite, even while contentious. I will delete comments which are ad hominem, as well as those needlessly profane beyond the realm of sputtering incredulity in reaction to some inanity, unless attributed to a quote.

Links to other sources are fine so long as they further the argument or expand on the discussion. All such comments and links are the responsibility of the commenter, and the mere presence herein does not necessarily constitute my agreement.

I will also delete all comments that link to a commercial site.