Polls, Illusions, and Bad Analysis

Gallop and Bloomberg polling tell us that tax reform is not an important priority to the American people.  CNBC says, “Many Americans aren’t sure they even need tax reform.”  The Gallop poll, taken each April, shows that only about 50% of the people polled believe that taxes are too high, while only 4% of the people Bloomberg polled believes it is a high priority.

Since individual tax reform is supposed to lower tax rates and make the tax system simpler, why would people show a lack of interest?  The poll results seem to defy common sense.   The answer is contained in the fact that the top 40% of tax payers pay 97% of federal taxes while 45% pay no federal income taxes.  A 2016 Tax Foundation report on compliance concludes “the latest official estimates of the eye-popping amount of time and money that Americans lose each year in complying with IRS paperwork—8.9 billion hours and $409 billion in lost productivity—indicate that the most important benefit of tax simplification may be the gift of time.”

Since 45% pay no taxes, the way the poll was conducted and the structure of the questions may have foreordained the results.  Random sampling is a fundamental principle of statistics.  However, in the case of the Gallop poll, random sampling of sufficient size would clearly bias the results since almost half of those polled would have been in the pay no taxes category.  Gallop never connected its results with the skewed distribution of individual tax payments.  In addition, in explaining polling results that fail the smell test, pollsters blame “the growth of cellphones and the decline in people willing to answer surveys.”  In 2014, the Huffington Post ran an article by Frank Wall, Severe Flaws in Polls.  In it he said, “The basic problem with polls is that public sentiment is quite like our body temperature; it goes up and down, somewhat randomly, and taking action too quickly in response to changes can lead to serious unintended consequences. … It is also increasingly clear that many polls are tainted by respondents lying to pollsters because they believe they cannot trust where their answers might end up …  That can lead to unintended consequences of self-fulfilling promises because misinformed poll results obviously do not tell the real story … .” Amen to that!

Michael Traugott, a University of Michigan political science professor who specializes in polling and opinion surveys got to the heart of bad polling very succinctly, “Put simply, if you ask the right sampling of people the wrong thing … you’ll get a bad result”.  Traugoot could have also added, you also get bad results, if you don’t dig deep enough into the polling to get to the real meaning or if you cherry pick results to support an agenda.  Bloomberg and CNBC which wrote about the Gallop results could be guilty of both.  Neither were balanced or insightful in their articles.

In Daniel Boorstin’s The Image, he says of public opinion that it “became itself a kind of pseudo-event, forced into existence for the primary purpose of being reported.  Expressions of public opinion became among the most powerful, the most interesting and the most mysterious of pseudo-events.”  Others like Darrel Huff who wrote How to Lie With Statistics and John Brignell who wrote Sorry, Wrong Number provide us a way to identify attempts to bamboozle and poor methodologies and analysis.

Gallop should be especially sensitive to its polling methodologies and analysis since it gave us President Thomas Dewey and more recently led Mitt Romney to believe that he was going to win the presidency.

Stories based on polling results probably should carry a warning, “reader beware.”

 

 

 

 

 

Author: billo38@icloud.com

Founder and president of Solutions Consulting which focuses on public policy issues, strategic planning, and strategic communications.

Leave a Reply