|
Electoral votes based on recent polls |
The
U.S. Presidential Election is just a month away, and if
you’re at all curious about how the election may go, you’re in luck, just turn
your attention to the latest political polls. You can compare the national approval ratings
for president Obama and (former) Governor Romney, how favorably they’re viewed
by “likely” voters, and who’s leading who in swing states like Ohio, Virginia,
and Florida. You can note which has the advantage on issues like unemployment,
foreign affairs, and government debt. You can survey how the candidates compare
among young voters, retirees, minorities, and women. You can follow how the
race is tracking on a day-to-day basis. If you’re interested in just about any
measure of how Americans may vote on November 6, there’s likely a poll for it. They’re
all just a Google search away, knock yourself out.
Political polls are everywhere in the run-up to every major
election, they’re quoted by the media, consulted by the campaigns, and, they’re
bad.
To explain why this is, I’d ask that you first consider a
question: Why should you, a voter deciding on how to cast your ballot come
Election Day, care at all about political polls?
Polls Aren’t Science
Pollsters, and there’s a bunch,
1 will tell you that
their polls are scientific and accurate. This is not true. Their argument is
that their methodology ensures that a statistically representative sample of
voters is used in every poll. The overall number of responders, the political
leanings of those responders, the questions asked, etc., are carefully calibrated
to give an outcome that is accurate within a small margin of error. This gives the
sheen of science to the whole effort, as if polls are just another demographic
study based on heaps of concrete data. They’re not. It’s certainly true that over
the decades pollsters have learned how to better sample the population. Gone
are the days when polls showed Alf Landon
2 beating FDR. But at the end of the day,
polls are based on the responses of people, and when it comes to people and
politics, you can throw science out the window.
Consider the emphasis in the following question: “Do you
agree that President Obama has done a poor job in addressing unemployment?” Admittedly,
this is a very simplified example of obvious bias, which is supposed to be scrubbed
from all modern poll questions. An unbiased question would better read: “Do you
approve or disapprove of the way President Obama is handling the economy?” But here
is an actual question from
a Washington Post poll from late last month: “Do you
think the federal government should or should not pursue policies that try to
reduce the gap between wealthy and less well-off Americans?” Would you call
this question unbiased? Maybe you think it is. Or maybe you consider terms like
“reduce the gap”, “wealthy”, and “less well-off” as loaded and apt to nudge
responders in a certain direction. What if the question were reworded this way:
“Do you think the federal government should or should not pursue policies that
try to shift wealth from those Americans better-off to those less well–off?” Which
wording is more biased and are they likely to elicit different responses? This
shows you just how hard, and maybe impossible, it is to actually extract bias
from any political discussion. If you can’t take bias out of the questions, how
can you have an unbiased poll?
Then there is sampling and response bias. Most polls are
conducted by calling people with landline phones, which is becoming an
increasing anachronistic approach in the era of mobile communications. Consider
the constantly shifting demographics (age, sex, economic background) of people
that use landlines as opposed to cell phones. What groups are over- or under-represented
in surveys conducted in this manner? Further, when are the surveys conducted,
during what days and at what times (i.e., who is home when the calls are
placed?)? How might this affect the bias of the results?
With response bias people may answer in a manner contrary to
what they believe or refuse to participate at all. After all, how honest are
people when talking to strangers about politics, a sensitive subject for many? What
type of person is willing (and available) to participate? How representative is
that person, or that aggregated group of people, of the voting population at
large? These questions aren’t easy to answer or dismiss.
Here’s a good example of how these factors can combine for
bad polling. Back in 2004, pre-election polling showed a very close race in Virginia
between President Bush and Senator John Kerry. And this seemed to be confirmed
on Election Day when exit polls indicated that Kerry was performing very well.
Yet when the actual votes were counted, Bush led Kerry by a wide margin at all
times (
he won comfortably 54% to 46%). The networks didn’t call Virginia in Bush’s
favor for many hours after the voting ended based on the strength of the inaccurate
pre-election and exit polls.
3
Polls are bandied about as accurate and unbiased. In other
words as a useful indicators of how the public is likely to vote. But they’re
often neither accurate nor unbiased. What’s the practical difference between a
bad poll and the daily political spin issued by a campaign? Intentional or not,
aren’t they both forms of misinformation?
Politics is Not a
Spectator Sport
It’s hard not to see the same relationship between political
polls and politics as we see between sports and sports statistics. Professional
and college sports are one of the tent poles of the vast and growing American
entertainment complex, and statistics are the drug of sports enthusiasts;
4
the careful tracking of performance, the rankings, the orderly measure of
players, teams, and leagues. There’s long been a cottage industry built around
baseball statistical research; fantasy football, which is all about statistics,
is one of the
most popular recreational activities in America; the essence of
entire sports are based on standings and rankings, and a player’s worth is
determined in hard data. As a nation of spectators we love to watch sports –
you could probably argue that the next Superbowl will be a more watched event
than the upcoming election – and statistics give us more to talk about and
discuss. It seems weightier to parse a team’s statistics and analyze
performance based on numbers than to simply describe and appreciate the physical
competition. It’s the data-science companion to the physical action-art.
For some, politics is the sport of choice. But of course
politics isn’t a sport, it’s not entertainment. At its core politics is about
how as a society we choose to live together, and it involves complex, convolved
issues. Issues that are hard to fully understand yet have an important effect
on everyday life. It’s hard to understand the current tax structure and the
implications of changes to the tax code; health care is a confusing tangle of
doctors and medicine, hospitals and insurance, regulations and paperwork; unemployment,
gay marriage, abortion, education, government debt, the European financial
crisis, war in the Middle East, they all dominate the headlines but none of
them have easy solutions. The real societal issues of the day, the issues that make
up the political landscape, all require strenuous discussion and wearisome compromise.
It’s hard, not fun; it’s tedious, not exciting.