Document Sample
election Powered By Docstoc
					Predicting Elections
In 1992 Bill Clinton and Bob
Dole ran for President.
All the TV networks and both
campaigns knew by noon
Central Time that Clinton had
won the election.
The networks waited to
broadcast this information
until 8pm Central Time, when
polls in the majority of states
had closed.
In every election since 1964—
except for 2000—the networks
knew the outcome before the
time the evening news was
Why is it that TV networks are
almost always right when they
predict elections?

1. They are careful to take a
   large, good sample (in
   most election years, the
   networks pool their
   resources to take a
   stratified random sample of
   as many as 10,000 voters.)
2. They repeatedly poll each
   state before the election,
   looking for consistency in
   the results. They use the
   results to pre-sort states by
   how they expect them to
3. They compute an estimate
   of each candidate’s
   percentage of the vote,
   including a margin of
   error, and only call states
   where the victory margin
   victory exceeds the margin
   of error.
4. The mathematics of
   statistics gives them
   confidence that their
   results will be correct
   between 95 and 99% of
   the time.
… So what went wrong in
Florida in 2000?
The polls from all news
agencies showed that the
Florida race (and the
nationwide electoral vote) was
too close to call—which was,
in fact true.
Most polls said Al Gore had
won Florida, but by a smaller
margin of victory than the
margin of error for the poll.
ABC and CBS said later they
felt pressure from advertisers
to make a decision, so they
called Florida for Gore around
Fox News called Florida for
Bush around midnight. (Later
analysis showed Fox’s poll
was also too close to call.)
After the Fox call, ABC and CBS
took back their call for Gore and
declared Florida too close to call.
Later reviews of the polls
showed the ABC/CBS poll
(which was also used by NBC
and CNN) had a higher
percentage of minorities than
had actually voted in Florida.
The Fox poll had a
disproportionately high
percentage of voters in the
northern part of the state,
which went heavily for Bush.
Several independent
investigations in 2001 and
2002 had mixed results.
Using a variety of re-count
procedures, four
investigations showed Gore
would have won Florida, while
three showed Bush would
have won the state.
However the Supreme Court
stopped the re-count, and
Gore conceded the election
long before these
investigations happened.
… So, basically the
networks were right. They
just too quick on the draw.

… It really was too close to

Shared By: