No Repeat of Election Night 2000

CBS News' Mason is intent on not making the same mistakes

CBS News Vice President Linda Mason hasn't forgotten the
debacle of Election Night 2000 and knows the White House race could be every
bit as tight this time around. She remembers how Voters News Service
(VNS)—the vote-counting consortium formed by ABC, CBS, NBC, Fox News Channel,
CNN and the Associated Press—greatly underestimated Florida's absentee
ballot, which prompted the networks to call the state for Gore shortly before 8
p.m. on Election Night. When the errors become evident by 10 p.m., the networks
admitted they'd been premature. More blunders by VNS led Fox News to call
Florida for Bush at 2:16 a.m. The other networks followed Fox's lead shortly
after, only to be forced to later declare the race too close to call. The
polling disaster kicked off a month-long electoral crisis and forced the
networks to disband VNS. In its place, the networks and AP have created the
National Election Pool (NEP), which they claim is better prepared for a close
B&C's Bill McConnell discussed changes in
Election Night prognostication with Mason, CBS' representative at NEP. She led
the network's investigation into the 2000 election meltdown and authored an
87-page report detailing the mistakes. She knows this election seems likely to
be just as close as the last one.

How likely are the networks to blow their calls
this time around?

I'm extremely confident we won't have the same troubles. VNS models were
all wrong, and their computers broke down. We've tested the new system
thoroughly during the primaries, and we've done hours of simulated stress
tests, overloading the system to make sure computers won't break down.

What's different besides better

First thing we've changed is, we've split the jobs VNS did in 2000
between two organizations. VNS computers were put under tremendous stress. Now
AP will have responsibility for collecting official tabulated results from
election officials. Edison Media Research will be doing the exit polling. This
system is much more robust because the two tasks will be checking on each

Does this mean competing news organizations will
be calling different states at the same time?

No. To make projections, all the networks are getting the same data. NEP
flatly says no one will call a race in any state before its scheduled close
but, once a state's polls close, every network will make its own decision on
when to call a race. But you'll see differences between the networks'
predictions. At CBS, we have a decision desk where our analysts have worked out
models used to make our own predictions. During the Iowa primary, CBS was first
calling for Kerry. In Wisconsin, CBS was dead last calling that state for Kerry
because Madison and Milwaukee results had not been received. I thought it was a
dangerous thing to do. Everybody else called it earlier. It's great that they
we're right, but you want to know why you're right.

What else is CBS doing differently?

We've brought our decision desk into the newsroom. In 2000, we were on a
separate floor analyzing the returns. We knew there were problems in Florida,
but that didn't get communicated to the newsroom.

We will now have a correspondent assigned to the decision desk. If the
same problems occur this year, the newsroom will learn about them immediately,
and so will viewers. Throughout the evening, we plan to be transparent. Viewers
will be along with us in knowing what's happening with vote counts rather than
CBS appearing omniscient, as we have in the past. If a counting error leads us
to restate a vote, we'll tell viewers why.

But better transparency is not enough. We'll also make it clear when our
predictions are based on exit-poll data and our estimates. Viewers will be
constantly reminded that mistakes can result. After investigating last
election's problems, I learned to my shock there are 2 million uncounted votes
in national elections. There are malfunctioning machines and lost bags of
votes. All kinds of things happen in a close election, and those votes mean a
lot. Every deviation is going to have an effect.

We've also improved how we deal with statistical problems encountered in
2000. We feel we've developed more-appropriate estimates of sampling errors. We
don't rely just on exit polls. We will do all we can to report actual returns
as soon as possible. We've added sample precincts statistically chosen. You
need to see the raw vote to see what's exactly happened. It varies from state
to state when tabulated results become available; some are released right away,
and some take longer.

If the election is too close to call even after
all the polls close, will you admit it or take your best guess to predict a

If it's very close—and there's every indication it will be—one thing
we learned in 2000 is that many small things have the power to sway a state one
way or another. There's no pressure from Dan or Andy [Anchor Dan Rather or News
Division President Andrew Heyward] to be first to call a state. The pressure is
to be right.