Page 428 - Battleground The Media Volume 1 and 2

Publ c Op n on: Are Polls Democrat c? | 0

as a symbol of political superficiality. For while Bryce’s vision may have been a
little naive, opinion polls can have a democratic function and provide a check
on political elites. Indeed, there is a case for us taking polls more seriously than
we do now.

ThE PoinT oF PoLLing
There has been a great deal of attention paid to the statistical shortcomings of
polls. This is, in part, because they are often associated with predicting electoral
outcomes: a complex matter that requires sophisticated sampling techniques
(in order to predict who will actually vote, who is lying, etc.) and where the
margin of victory may be less than the margin of error. In this spirit, scholars
have often chided the news media for their failure to report or understand the
technical aspects of polls.
If the debate about the ability of polls to forecast elections often takes center
stage, we are better off seeing it as a trivial sideshow. This use of polls has very
little to do with exploring or representing public opinion: their purpose is to pro-
vide a commentary on the electoral race. From a democratic point of view, this is
an ultimately pointless exercise that might be done just as well by bookmakers.
Moreover, an obsession with accuracy is misplaced. Methods matter of
course: there is no shortage of surveys based on ad hoc or self-selecting samples
that cannot claim to be representative. But most opinion polling takes care to
use sampling techniques designed to represent a broadly representative cross-
section of the public. And while polls can never predict how people will vote
with pinpoint accuracy, sampling techniques allow us to get a flavor of pub-
lic attitudes using surprisingly small samples. We should not lose sight of how
remarkable it is to be able to get a sense of what the population is thinking by
talking to only a tiny fraction of that population.
But sampling, almost by definition, is not an exact science. Its statistical
method is based on notions of probability rather than certainty. A good poll
may claim to be accurate within two or three percentage points, but it does so
on the basis of a level of probability (say, 95 percent). A poll that is a few per-
centage points out in predicting an election result is not so much “wrong” as
misinterpreted.
We are, perhaps, seduced by the neat numerical precision that polls can pro-
vide. What is useful about polls, however, is their ability to suggest patterns and
tendencies rather than their exactitude. Once we recognize this, we can see how
the more profound limits of polls have less to do with science and more to do
with the nature of the artifice. To put it another way, the issue is less who is
asked, than what they are asked, in what context, and what we make of their
responses.

ThE ProDuCTion oF PoLLs

There are three issues here. First, as many critics have pointed out, there is
nothing authentic about the poll versions of public opinion (or, for that matter,