Virtually the entire public sense of the presidential race so far is based
on opinion polls. I have been critical of most of these polls, even some
of the most well-known ones, because of their highly subjective
"weighting" of raw data, their statistical inaccuracy and methods, and
the inherent inability of any objective assessment of them.
I realize that the rejoinder to my criticism of the dependence on polls
is: What is the alternative? Fair enough, if you are requiring some
pseudo-mathematical support for assessing who is ahead, who is behind,
and by how much.
I have long asserted that there are two kinds of polls which are genuinely
credible. First, most polls privately taken for candidates, and paid for by
them (usually at relatively high cost) are reasonably accurate. They often
do not reveal good news for their candidates, but that is not their purpose.
Their purpose is to instruct campaigns on how their candidates are doing
with voters and on specific issues. Private pollsters are usually
professionals, and know how to get reasonably accurate information from
their work --- at a price. They also know that if their work does not produce
accuracy, they will not be hired in the future.
Public polls, on the other hand, are often distorted by various factors of
economic cost, subjective judgments about how to "weigh" the raw data,
their size, and chronic ambiguities in how questions are asked. Most of
these polls are produced by media organizations or polling firms which
serve one political party or another, or one ideological view or another.
The second type of political polls which are usually credible are major
polls that are taken just before an election. This means polls done between
two days and one week before an election. No professional pollster could
last very long if their polls in this time frame were routinely very wrong.
The polls most of us read, or have access to, at this stage of a presidential
campaign fit neither of these conditions. The polls we do read have
relatively small samples, are weighted by wildly various assumptions
(including weighting how many Democrats, Republicans, and independents
to count from the raw data), are often worded so as to "tilt" the replies, and
face historically low numbers of responses from initially selected random
voters (a factor which distorts statistical accuracy significantly more than
most pollsters will admit to), and so far often do not distinguish between
"registered" and "likely' voters. I would contend that a poll of registered
voters is far less useful than one which uses "likely" voters.
Public polls so far indicate that the presidential race is very close. It may
be, so but we have no real evidence, in my opinion, that it is close. More
importantly, we have no evidence at all that it WILL BE close in November.
The common sense statement that the election will be determined by how
voters will feel about their own economic circumstances on election day,
or on the day they make up their mind who they will vote for, probably has
more validity than any other judgment about who will win and who will lose.
We are now coming up to having only 100 days until election day, but the
economy, unemployment, manufacturing and housing numbers remain
static, as they have virtually through the presidential administration of
If this continues, or is only slightly moderated, he will lose. If the economy
dramatically turns around to the upside, he will win. As we draw closer to
election day, the public polls will more and more resemble the private
Whether the reader likes it or not, public polls today tell hm or her relatively
little about how the presidential race is going (other than the obvious that
most Democrats will vote for Obama, and most Republicans will vote for
Romney). Even if the reader has access to the best private polls, they only
know something of current opinion.
A great deal of gamesmanship is now taking place, and will further intensify
after the national conventions. The purpose, on both sides of course, is to
influence the result in November. There will be much of this, and events (now
unpredictable) will have impact no poll can predict.
I want to say one more time that there is only one preeminent question in the
presidential election of 2012:
"Does the majority of American voters, expressed through the electoral
college system established by the U.S.. constitution, want four more years
of Barack Obama, or not?
Copyright (c) 2012 by Barry Casselman. All rights reserved.