Joan Shorenstein Center on the Press, Politics and Public…

       Joan Shorenstein Center
on the Press, Politics and Public Policy

              Discussion Paper Series



                  Exit Polls:
   Better or Worse Since the 2000 Election?

                     By Robin Sproul
 Vice President and Washington Bureau Chief, ABC News
        Kalb Fellow, Shorenstein Center, Fall 2007


                                #D-42




  © 2008 President and Fellows of Harvard College. All rights reserved.
Introduction
There have been so many problems with exit polls in the last four national
elections that news organizations approach 2008 election night coverage
without a great deal of confidence in what those polls will show. The six news
organizations that jointly conduct exit polls, ABC, CBS, CNN, Fox, NBC, and
the Associated Press have been on a roller coaster ride since Election Day
2000, with a great many successes, some spectacular failures, enormous efforts
to "fix" the polls, and millions of dollars spent in the process.


This paper will review the concerns with exit polls identified by the news
organizations after the 2000 election mistakes, and update progress made since
that time. It will also look at new concerns about exit polling raised since the
2000 election.


History
Throughout American history, journalists have covered elections. The
American public learns election results from a free press in as timely a manner
as the era's news delivery system allows. Today, the news cycle is one
measured in instants, and the public expects rapid and accurate information
"now." In the last four decades, exit polls have been a crucial and for the most
part very accurate part of providing that information. Exit polls are surveys
conducted as voters exit their polling place on Election Day. Reaching voters
at that moment is important because it overcomes the problem pollsters have
conducting telephone polls: people tend to misreport whether they voted or
not. The "who won and why did they win" coverage on election night comes
mainly from exit poll results, with the information reaching most Americans by




                                         1
television, and increasingly by the internet. The development and evolution of
exit polls is a story spanning 40-plus years.


In the summer of 1964, ABC, CBS, and NBC, along with the AP and UPI wire
services jointly formed an organization called the News Election Service (NES)
to pool resources to gather vote count information. By joining forces, the
news organizations hoped to get more accurate vote totals nationwide than any
one organization could get on its own. The calling of races and the
interpretation of election results, though, were done separately by the individual
news organizations.


In 1967, Warren Mitofsky, widely known as the "father of exit polling," began
conducting them for CBS News. Advances in computer technology enabled
the quick analysis of large amounts of data, and the competition to have the
best election night coverage drove the development of the methodology. Exit
polls collect information gathered from large numbers of citizens (more than
100,000) as they leave their polling place on Election Day. Most importantly,
exit polls ask for whom the voters cast their ballot, but they also gather
demographic information to determine whether differences in such things as
income, age, race, gender, and education impacted voting patterns. Typically,
the exit poll also questions the voter's position on issues that were important in
the race. On election night, analysts tell the story of the election using all of
this information, which is broken down state by state, critical to telling the
electoral college story, and nationally.


By 1980, with networks conducting individual exit polls, results were being
used not only for interpreting election results, but also for competitively


                                           2
projecting races. That year, NBC News announced that Ronald Reagan had
won the presidency three hours before some polls had closed on the west
coast, resulting in an early concession speech by Jimmy Carter. This was the
first great exit poll controversy. Critics argued that the early call depressed
turnout in the west. Exit polls were criticized by Congress and hearings were
held. The networks agreed not to report results until most polls were
scheduled to close in any given state.


In 1984, however, all three networks used exit poll results to declare Ronald
Reagan the winner over Walter Mondale before polls had closed in the west.
Congress then passed a resolution urging broadcasters to voluntarily refrain
from characterizing or projecting the results of an election before all polls for
the office had been closed. Taken literally, that would mean no presidential
winner could be declared until polls had closed in all states, including Alaska
and Hawaii. Even in a landslide election, networks would have to sit on the
news until well after 11:00 p.m. ET. In the age of "instant news" and the
internet, that is a scenario difficult to imagine. In practice, the networks
continued to follow a state-by-state policy, calling races when most of the polls
in a given state were closed.


In 1985, the presidents of ABC News, CBS News, and NBC News were called
to testify about exit polls and projections before the House Subcommittee on
Elections. At that time, Congress was exploring the idea of adopting a uniform
poll closing bill, in which all polls in the continental United States would close
simultaneously. NBC's Lawrence Grossman, who testified that day, says the
news chiefs made a promise not to "project or characterize" results in any state
until after its polls closed. Grossman wrote in 2000: "Our promise to Congress


                                         3
was a mistake that continues to haunt television's coverage to this day."
Grossman adds:


      On any given election day, anyone who listens to what reporters,
      analysts, anchors, and campaign staffs say on the air can figure out
      well before the polls close who's ahead, who's behind, and how
      close a race is. The only way not to get an early peek at the voting
      trends and results is not to turn on any television, radio, or
      computer.1

In 1990, ABC, CBS, CNN and NBC, in an effort not only to save money, but
also to address public concerns about the ultra-competitive nature of election
projections, formed Voter Research and Surveys (VRS). This consortium
conducted joint exit polls for the 1990 and 1992 elections, and provided the
networks with simultaneous projections. With this pooling of information,
access to the early waves of incomplete exit poll data became a hot political
commodity on Election Day. Political reporters traded information with
political insiders all day. Although the partially weighted data is not meant for
public consumption, during the 1990s, most of the people who saw the early
data knew its limitations.


In 1993, NES was folded into VRS, and the Associated Press became a partner
with the networks in the newly formed organization, Voter News Service
(VNS). Fox News became a VNS member as of the 1996 election.
VNS provided both exit polls and vote counts. But the competitive nature of
election night coverage again became an issue, when ABC News formed its
own decision desk and was able to call races significantly ahead of the VNS
projections in the 1994 election. Subsequently, each network developed its
own "decision desk" to call elections, separate from the VNS projections. The


                                        4
network projections were based on statistical models that combined historical
information about past voting in sample precincts, the exit poll data, actual vote
count, pre-election polls, and in-house election experts.


Over the first 30 years, the exit polls had a reputation for accuracy in projecting
election results. During the 1990s, VNS had excellent results: only one error in
700 projections.2 It is important to remember, though, that the raw data from
exit polls has always required careful weighting with actual vote totals and
comparison models. The data streaming in as Election Day progresses has
never been "accurate" until it is weighted. Before internet sites started
releasing leaked partially weighted data, leaving the public with the impression
that the numbers were "off," exit polls had a very good reputation.


Historically, exit polls have also provided deep and reliable analysis of election
results. Academics depend on the data to understand the changing American
electorate. Political stakeholders use the data as they develop strategies and
policies, attempting to understand public opinion. Exit polls help journalists
frame the meaning of election results as early as the night of the election.
While winning and losing campaigns offer their preferred explanation of the
cause of their victory or defeat, journalists use exit poll analysis to explain who
voted for whom and why. In this way, the "election mandate" discussion
begins not with self-serving politicians, but with non-partisan analysts
characterizing results based on the large data sample provided by exit polls.
The value of the timeliness of the exit poll information, particularly in the
speeded-up news cycle, cannot be under-estimated. Americans form opinions
about the "why" of a given election very quickly, and these first impressions are
apt to last a long time.


                                         5
There are other studies done of issues of importance to voters, but it is the size
of the sample, the timeliness factor, and the fact that those interviewed just
finished voting that makes exit poll data so highly valued. Plus, the exit polls
present data gathered in all 50 states and the District of Columbia, every
national Election Day, providing highly specific regional breakdowns of voting
trends over time.


The 2000 Election
Things started going badly for exit polling in the 2000 election. In such a close
presidential race, with Florida being a decisive state, VNS made a bad call.
Even before all the polls had closed in the state, VNS called Florida for Al
Gore at 7:52 p.m. The networks and the AP made the same call, all at about
the same time. Some two hours later, the call was withdrawn. At
approximately 2:00 a.m., the networks (but not VNS and AP) called Florida for
George Bush, and that call was withdrawn within another two hours. The
election was simply too close to call.


The evening's mistakes cannot be blamed on exit polling alone. The problems
had as much to do with bad vote counts and bad information reported by local
election officials as they did with any problems from the VNS election models.
However, in the aftermath of the 2000 election problems, each television news
organization commissioned a report on its own election night errors. In those
reports, several concerns about exit polling were raised.




                                         6
Much of the internal criticism focused on how exit polls were done and how
projections were made. The conclusion of CNN's publicly released review
summarizes what the other networks found as well:


       The supposedly sophisticated system of polling is not nearly
       sophisticated enough. It is a flawed system that fails to take into
       full account many dynamic factors--absentee balloting, early
       voting, demographic change in key precincts, a declining response
       rate to polling generally, the quality of questionnaires, vote
       undercount, mistaken balloting, computer error, human error, and
       more.3

Recommendations for fixing the exit polls post-2000 included upgrading the
VNS computer system, reviewing the statistical models, conducting additional
research on the non-response and exit poll errors, studying the absentee voter
situation, and providing quality control in the vote counting operations.4


The networks and the Associated Press have continued exit polling in each
general election since 2000, and have worked hard to fix the problems
identified seven years ago. Pressure from the public and politicians in
particular to "get it right" has been constant. There have been calls to end exit
polling, mostly because of controversies over early projections and the 2000
election problems. Former presidents Jimmy Carter and Gerald Ford led a
bipartisan review of the 2000 elections, reporting to Congress on July 31, 2001.
The National Commission on Federal Election Reform was highly critical of
exit polls and made the following recommendation: "The Commission strongly
encourages citizens not to participate in exit polling."5




                                         7
Post-2000
The Computers and Technology
Although VNS overhauled its computers and custom software systems after
the 2000 election, at a cost of approximately ten million dollars, they simply
were not ready in time for the 2002 election. The exit polls were a disaster.
There were massive technical failures on election night, yielding no useable exit
poll data. Political scientist Michael Traugott, who has studied exit polls for
years, says there was another problem: the VNS statistical models were out of
date. These models of individual precincts, used to project the outcome of
races, were conceptually appropriate but did not use the latest statistical
theories and models available.6


By 2003, the pool members shut down VNS and formed the National Election
Pool (NEP), hiring Edison Media Research and Mitofsky International to
conduct the 2004 exit polls. Edison/Mitofsky (E/M) had to design, develop,
and implement a full exit poll and election projection system starting in January
2003, which would be ready in time for the 2004 primaries and election.
Although there were two technical disruptions on election night in 2004, those
problems did not impact the accuracy or delivery of the exit poll results. There
were no obvious technical problems in the 2006 election.


The Absentee/Early Vote
In reviewing their 2000 election coverage, the networks expressed concern
about the growing number of absentee/early voters nationwide and the
potential impact on the accuracy of future exit polls. The underestimation of
the 2000 Florida absentee vote numbers was a contributing factor to the exit
poll problems in that state. VNS had estimated the size of the absentee vote to


                                         8
be 7.5%, based on historical data. In reality, it turned out to be 12% of the
total Florida vote.7 Also, although VNS had correctly assumed there would be
a higher percentage of absentee Republican voters than Election Day
Republican voters, the estimation made was too low, probably because of a
significant Get Out the Early Vote (GOTEV) effort by the party. With such a
close election, the number and party breakdown of absentee voters were critical
factors in predicting the outcome of the election, and the exit poll model was
not accurate in estimating either.


What has happened since 2000 is that the quiet revolution in the way
Americans vote has continued. Without great attention being paid to the trend,
the country is slowly moving away from Election Day voting. Because election
administration is so decentralized in America, it is difficult to accurately track
the state-by-state and county-by-county voting rules, but there are some clear
trends that jeopardize the accuracy of future exit polls.


Alternatives to Election Day voting have been introduced with increasing
frequency, with the number of people who choose "convenience voting"
escalating. Methods of convenience voting are expanding, with the
liberalization of absentee ballot laws, the expansion of vote-by-mail elections,
and the growing number of in-person early voting options (hereafter, the term
"early voting" will generally refer to all of these voting options). This trend is
likely to continue as states and counties across the country look for ways to
encourage people to vote, reduce lines at polling stations, and save money on
Election Day. Voters have responded, not necessarily by voting at higher rates8
but by signing up for early voting options in greater numbers every year. In
nearly every state, the percentage of those who voted earlier than Election Day


                                         9
rose between the 2000 and 2004 general elections.9 In some states it was a
significant increase. In fact, by the 2004 presidential election, 26 million voters
opted to cast their ballots early. That represents more than a fifth
(approximately 22%) of those who voted nationwide.10


Looking at some key states and the most recent trends, the percentage of those
who voted absentee/early in California's general elections are telling:


       2000          24.5%
       2002          27.1%
       2004          32.6%
       2006          41.5%11

As many states are doing, California is making it easy to vote early. Residents
can now register as permanent absentee voters. Votes can be cast as early as 29
days before Election Day--necessitating the development of innovative
campaign strategies to cope with the rolling voting schedule.


Looking beyond California, there are a number of states in which early voting
has caught on in a big way. Here is a look at the ten other states with the
highest percentages of convenience voters in the 2004 general election:


       Oregon               100%
       Washington           67.8% (88.5% in 2006)
       Nevada               53.1%
       Texas                51.1%
       New Mexico           50.6%
       Colorado             47.8%
       Tennessee            47.3%
       Arizona              40.8%
       Florida              35.9%
       Arkansas             33.4%12

                                         10
If the trends continue, and they are likely to do so, these convenience voting
numbers will be even higher in 2008 and subsequent elections. What does this
mean for exit polls? Respected survey research expert Paul Lavrakas calls the
situation a "time bomb for 2008," saying "the early vote data has become just
as important as the exit poll data" in many states.13 In fact, exit polls are no
longer conducted in Oregon, where all voting is done by mail, or in
Washington State, because of the high level of early voters, according to Joe
Lenski of E/M.14


Whereas in previous decades, it was relatively easy to estimate any given state's
absentee/early voting outlook, the rapid rise in the number of these voters
from election cycle to election cycle makes it difficult to do that now. There is
simply not enough historical data against which to compare current results.


Again looking back at 2000 and the way VNS incorrectly estimated the number
and party breakdown of Florida's absentee voters, many questions are raised
for further study. How closely do early voters resemble Election Day voters,
and can assumptions be made about those similarities or differences? How
does that differ state to state? How does that vary from election cycle to
election cycle? Also, how do news stories that break just before Election Day,
after many early voters have cast their ballot, change the voting differences
between the Election Day voters and the early voters? Political scientist
Marion Just remembers that Ross Perot's assertions that Republicans were
planning to disrupt his daughter's wedding, a story that broke two weeks before
the election in 1992, caused a shift in support for Perot just before Election
Day.15




                                        11
Studies done by early voting expert Paul Gronke found that "early voting varies
in reasonable ways: voters who are willing to identify with a political party,
voters from areas with higher commute times, incomes, and average
educational levels tend to cast their ballots earlier."16 Gronke also cites several
studies showing that early voting options tend to attract the most politically
active voters. In some areas of the country, that is an advantage for
Republicans, and in some areas Democrats benefit.17 Other studies find little
partisan difference between early voters and Election Day voters.18 There is
evidence that older Americans are more likely than others to vote early.19


Although early voting systems themselves do not seem to benefit one party
over the other, regional party mobilization can impact shifts in the partisan
breakdown of those voters. For example, Democrats in Iowa focused turnout
efforts on early voters in 2006. At one point, Democrats outnumbered
Republicans five to one in requests for absentee ballots, although the parties
each have about the same number of registered voters in the state.20 These
party efforts to get their most ardent supporters to vote early can vary from
one election to the next, causing significant impact on the accuracy of early
voting estimations. The NEP will have to devote significant (and growing)
resources to pre-election telephone polling to get accurate information about
how and why the early votes were cast.


In a review of the 2004 exit poll results, Edison/Mitofsky recognized the need
to upgrade the telephone surveys measuring the early vote results. They
suggested that the size of telephone samples of early voters increase in future
elections, particularly in states where the absentee vote is a "large proportion"
of the total vote.21 Back in 2000, Joe Lenski and Warren Mitofsky were


                                        12
running the CNN/CBS decision desk on election night. When interviewed for
CNN's review of 2000 election coverage, they expressed concern about
mistakes caused by early voter levels far lower than the ones seen in 2004:


       Mitofsky and Lenski, believing the current VNS exit poll models
       are inaccurate in states where the absentee vote is more than 10
       per cent, also recommended that phone surveys be conducted in
       advance in states where the absentee vote is expected to be that
       high.22

As the 2008 election approaches, 30 states have early voter rates greater than
10%.23 Edison/Mitofsky is increasing efforts to account for these voters with
telephone surveys. Joe Lenski says the number of pre-election telephone polls
was increased for the 2006 election to cover ten states, and will be increased
again in 2008. However, as the number of pre-election telephone polls
increases, so does concern about the impact on the accuracy of exit poll
projection and analysis results.


For example, a primary reason that exit polls are so highly valued is that the
individuals surveyed have exited a polling place where they have just voted.
Their memory of how and why they voted is much more likely to be accurate
than it would be at any other time they are surveyed. And, since studies show
that voting is such a socially desirable behavior that people will tell pollsters
they voted when they did not, 24 exit polls all but eliminate the percentage of
those who would lie about actually voting. With the telephone polling
necessary to measure the early voting results, the uncertainties of memory and
actual voting behavior must be weighed into the results. As the percentages of
early voters go up, so do the percentages of those uncertainties.




                                         13
Telephone polling in general is a survey method with its own problems. Every
polling professional knows that the number of people willing to respond to
telephone polls has dropped lower and lower as Americans have been
overwhelmed by telemarketers and telephone solicitations. Answering
machines and Caller ID service have compounded the problem, with more
people refusing to answer calls from numbers they don't recognize. Pollsters
also have been dealing with the ever-increasing number of Americans who
have no landline telephones. Three out of every 20 American homes have no
landline telephone, though most of those homes have at least one working
wireless phone.25 Paul Lavrakas says that surveying persons reached on cell
phone numbers in the United States is a very complex undertaking. He says
that there are enormous challenges to doing it legally, ethically, and in ways that
gather the highest quality data. These telephone survey problems are escalating
just as the accuracy of exit polls will be more dependent on telephone polls to
measure those who vote early.


Another factor to consider has to do with the youngest voters, those most
likely to have only cell phones. In 2004, 4.3 million more 18- to 29-year-olds
voted than in 2000, a 9% increase in turnout (from 40 to 49%), and more than
twice the turnout increase for the overall electorate.26 Overall, they represented
17% of voters. In the 2006 midterm election, young voter turnout increased
for the second major election in a row with an increase of 3% over the 2002
midterms.27 As the youth vote numbers increase, they become a more
important subset of the electorate. And, there is some evidence that at least
those in college vote at a high absentee rate.28 While there is little information
about the overall absentee rate for the 18- to 29-year-old voters, the inability to
reach this growing group of voters who carry only cell phones will present


                                        14
ongoing challenges to the NEP. Despite the fact that research shows you can
weight poll data to compensate for not being able to reach the young voters
who have only cell phones, 29 as their numbers increase and as the numbers of
cell-only Americans increases in general, the challenges to the NEP grow
commensurately.


Summarizing, the need to supplement exit poll results with substantial
telephone polling in order to get as accurate a picture of the early voters as
possible is clear. For the NEP, however, this is an expensive (and complex)
undertaking at a time when financial pressures continue to rise.


Non-Response
In general, people are less inclined to respond to polls than they used to be.
This trend impacts all kinds of polling methodologies, but is evident in the
declining willingness to respond to exit polls as well. In their reviews of the
2000 election coverage, the news organizations expressed concern about the
declining response rates. Here's what the numbers look like:


      1992          60%
      1996          55%30
      2000          53.7%
      2004          53.2%
      2006          45% (non-presidential election)31

Warren Mitofsky noted that in the 1960s and 1970s, exit poll response rates
were in the 70-75% range. He speculated that the later decline in response rates
was due to many factors: the bad public reaction, starting in the 1980s, to early
projections on TV; the criticism by Congress of TV's early projections; the laws
in some areas that require exit poll interviewers to stand 50-300 feet back from


                                        15
the polling place; and the overall decline in response rates to all kinds of
surveys.32 Other factors may include the widespread criticism of exit polls after
the 2000 elections, and the public's continuing distrust of the media in
general.33


Even though the exit poll response rate continues to slowly decline, Dan
Merkle and Murray Edelman published research in 2002 that found little
relationship between response rates and survey error. Further, they found
surveys with higher response rates were not necessarily more accurate.34 So,
although the NEP members were concerned about declining response rates in
2000, and the rates have continued to decline, this is not a major problem in
and of itself.


Non-Response Bias
Although the response rate alone does not cause survey error, with a lower
response rate there is a correspondingly higher risk of non-response bias. If
the group of people who choose not to take the exit poll voted significantly
differently from the group that did choose to respond, the poll results will have
an overall "bias." An example: In 2004, the exit poll response rate for voters
age 60 and over was 43%, compared with a 55% response rate for younger
voters.35 That level of differential response would skew results if the data were
not corrected. Every effort is made to weight the data for the differences in
response rates that can be observed (such as age, race, and gender), but
differences that cannot be observed may not be fully accounted for until the
official vote returns are in.




                                        16
In the last two national elections, 2004 and 2006, there was a statistically
significant Democratic bias in the raw exit poll data gathered at the precinct
level. In simple terms, after the election, the estimates gathered by polling
place interviewers are compared to actual election results. In 2004, precinct
level exit poll results overstated the Kerry-Bush difference by 6.5 points in
favor of Kerry. In 2006, the exit poll's Democratic-Republican margin was
overstated in favor of the Democrats by 6.2 points.36 As exit poll data came in
on Election Day 2006, it was clear that Democrats were again being
"oversampled." That led at least one NEP member, Fox News, to announce
on air (9:25 p.m. ET) that they would not rely on exit poll numbers in their
projections.


The level of exit poll error in the last two elections is the highest since records
have been kept, though the overall tendency to over-sample Democrats has
been in evidence for the last several election cycles. The problem is seen
elsewhere as well. In 2006, a private company doing a statewide exit poll in
Wisconsin found significant Democratic bias in two of the three races they
were polling. The greater willingness of Democrats to participate in the
Wisconsin poll, and the higher than expected number of absentee voters
contributed to the problem there.


In the review of their 2004 exit poll results, Edison/Mitofsky reported the
primary reason for the bias was likely to be that more Democrats than
Republicans agreed to participate in the exit poll. E/M found no evidence of
problems in how data were processed, nor in the sample selection of the
polling places. In an interview, Warren Mitofsky blamed the problem on "the
failure of interviewers to follow the selection rate."37 Interviewers are


                                         17
supposed to follow a strict pattern of selecting voters, such as every third or
fifth voter, in order to get a randomly selected sample. But polling places can
be confusing and crowded, and interviewers can be pressured by those who
volunteer to take the survey, particularly in precincts where few people selected
agree to respond.


The Edison/Mitofsky report also mentioned several factors that may have
contributed to the bias in 2004, including distance restrictions imposed on
interviewers by local election officials, weather conditions, precinct
characteristics, questionnaire length, etc. However, E/M's chief
recommendation for improving the bias in future exit polls was better training
of the polling place interviewers. They committed to hiring fewer students and
young adults in the future, because older interviewers delivered better precinct
level results. They looked at other factors that correlated with skewed response
data including the race and party affiliation of the interviewer.


It is still unclear exactly what caused the Democratic bias in the 2004 exit poll
data, but improvements made for the 2006 election did not eliminate
Democratic over-sampling. Joe Lenski of E/M said there was better training
of interviewers before the 2006 election, and some on-site supervisors were
hired. Efforts to eliminate the poll bias in 2008 continue to focus on the
interviewer selection and training, and what Michael Traugott calls the
"logistical nightmare of hiring and training"38 this very large temporary staff for
a single day's work every two years.


Stanford University political scientist Jon Krosnick is an expert in survey
methodology. He has another suggestion for tackling the Democratic bias.


                                        18
Having done extensive research on how ballot name placement affects vote
choice, Krosnick notes that the Democratic presidential name is always listed
first on the exit poll questionnaire. He suggests that name placement be
varied.39


It is important to note that the exit poll bias existed in the raw data, but did not
cause any incorrect projections to be made. Once vote count totals were
merged into the exit poll results and the data were weighted to correct for over-
sampling, the results proved accurate. However, in order for there to be
confidence in future exit polls, the bias must be accounted for and corrected.


One final point about exit poll bias that has been raised by several academics
and Pollster.com's Mark Blumenthal involves transparency. The news
organizations running exit polls for the past four decades have had a policy of
confidentiality in the precinct-level data. In order to protect the privacy of
voters who participate in the surveys, a certain level of specificity in the data is
never released. Some academics think that the NEP could "blur" precinct
locations enough to protect voter privacy, thereby allowing scholars full access
to archives of raw data. Michael Traugott and others explain that's not what's
happening now:


       The data that are deposited are weighted in a complex way that
       accounts for some nonresponse adjustments, the demographic
       characteristics of the respondents, and, most importantly, to the
       actual outcome of the election... When the outcome of past
       elections was not so close and the introduction of new voting
       technology was less common, this was a satisfactory procedure.
       But in the context of the leaks in 2004, the competing theories
       about why and how this happened, and the quality of the exit poll
       results, data weighted to the actual outcome of the votes was no


                                         19
       longer a satisfactory dataset for many consumers of exit poll
       results.40

Particularly because of the issues of bias, it is possible that full disclosure would
give the public and academic researchers more confidence in the data. With so
many citizens concerned about election integrity since the 2000 election
problems, those who conduct exit polls cannot appear to be "hiding" data.
Pew Research Center president Andy Kohut points out that the process of
voting has become more politicized, with citizens challenging the accuracy of
official vote counts, and with bloggers insinuating voting fraud in recent
elections.41 Views of the media have also become polarized, and any
perception that the media are hiding data makes the exit polls appear suspect,
even when they are not.


Leaking of Exit Poll Data
The leaking of early and only partially weighted exit poll numbers in 2004
caused much public confusion. Those numbers appeared on multiple internet
sites on Election Day, and they indicated a strong showing for John Kerry.
Some sites crashed because of the heavy traffic, with the Drudge Report, Slate,
and others seeing huge increases in visitors. Early exit poll data had leaked
onto the internet in 1996 and 2000, without much public notice. In 2004, the
prevalence of political blogs and increased web traffic caused the data to move
among sites at lightening speed, reaching millions of people. Most of those
people had little contextual understanding of the data. Richard Morin wrote in
the Washington Post:


       If a few hours on the roller coaster of ecstasy and agony were all
       that anyone had to endure, only the political junkies would be
       interested in the whys and wherefores of the exit poll confusion.

                                         20
      But the false picture had real impact: The stock market
      plummeted nearly 100 points in the last two hours of trading, and
      the evening news was replete with veiled hints of good news to
      come for the Kerry campaign.42

Even today, a vocal minority believes those early waves of exit poll data in 2004
were "correct" and Kerry actually won, despite the final results of the election.


The NEP was so concerned about the leaks in 2004 that they put a strict
quarantine in place for Election Day 2006. Only a handful of officials from the
NEP saw the early, partially weighted data. The quarantine worked, and will be
used again in the 2008 election. Even though in 2006 the public did not see the
over-sampling of Democrats in the early data, it was there. No one in the NEP
is comfortable knowing that the precinct level problem has yet to be solved.
Paul Lavrakas served as a quarantine room monitor in 2006, watching the early
waves of data come in: "I experienced firsthand how precarious is the
confidence of the NEP media sponsors in the validity of the exit poll data."43
Mark Blumenthal44 got a call from "someone on the inside" of the NEP saying
"do not trust any exit poll data tonight."


The Single Source Problem
Before the networks and the AP joined forces to conduct a pooled exit poll in
1990, networks conducted individual exit polls. Network and print journalists
wrote stories analyzing election results by looking at more than one stream of
data. No journalist likes having a single source of information on a news story,
but news organizations can no longer afford to run parallel exit polls for
purposes of comparison. The public would clearly be better served with
multiple exit polls, because when single-source data are wrong, there is limited
opportunity for correction. Marion Just calls the current situation "an insidious

                                        21
problem that over weights this single stream of data because every news outlet
uses the same story."


It's easy to see how much of an editorial problem is caused when exit poll
analysis data are "off" in the first news cycle. One example of how a second
source of data would have contributed to a better understanding of election
results happened in the demographics area of the 2004 exit poll. On election
night and in subsequent news cycles, media organizations reported that
President Bush received 44% of the Hispanic vote, citing the only source
available, national exit poll data. This result surprised many observers because
it was a large and unexpected increase over the president's 2000 support from
Hispanics. The story got a lot of attention. Two weeks later, the exit poll data
were corrected to reflect final election results. The difference was significant:


       Bush ­ Kerry         44 ­ 53       (Exit Poll ­ election night)

       Bush ­ Kerry         40 ­ 58       (Exit Poll ­ corrected results)45

One reason for the change was that E/M had incorrectly weighted the absentee
vote, particularly in Texas, which has a high number of Hispanic voters. John
Harris of the Washington Post wrote about confusion over the numbers a month
after the election:


       Political analysts are still scratching their heads over what share of
       the crucial Hispanic vote President Bush won last month. ...
       News media exit polls on election night reported Bush winning
       44% of Hispanics this year, a startling nine percentage-point jump
       from 2000. Some skeptics weren't buying it, saying the data were
       flawed.46




                                        22
The combination of having only one source on election night and the ongoing
problem accounting for the early voters caused incorrect demographic
information to be reported on election night.


Some academics have called for an additional, independent national exit poll to
be conducted.47 This is a scenario that is highly unlikely because of the cost of
such an undertaking. The Los Angeles Times, which has run its own national exit
poll since 1980, recently announced that for financial reasons they would no
longer be doing that. Director of the Harvard Opinion Research Program
Robert Blendon thinks the NEP news organizations should rely on more
extensive and competitive pre-election surveys to put together as full a picture
as possible in trying to understand election results and voters' motivations, so
they are not as reliant on exit polls election night.


In the 2000 election reviews, the networks also identified having limited
sources of vote tabulation as being an issue of concern. But in 2000, there
were still two sources of vote count: VNS and the Associated Press. The CBS
News review said that if they had checked both of those sources in 2000, the
second incorrect call of the evening, the call for Bush at 2:17 a.m., would not
have been made. The AP's vote count was corrected one minute before CBS
News made the call, but they had not seen the correction, relying only on the
VNS count.48


The vote count situation now is worse than it was in 2000. There have not
been two streams of vote count since the 2002 election. AP now provides the
sole source of vote count data, although some individual state websites are
becoming more reliable with back-up information. Even though the AP's vote


                                         23
count has been reliable, accurate, and has built-in redundancies, it is still
another single source of key information on election night.


Finally, having a single source of data on election night means that there is no
back-up if and when the system crashes. That is exactly what happened in
2002, when technical failures overwhelmed the exit poll system.


The Questionnaire
The exit poll questionnaire is written by a committee of NEP members. Many
academics, including Robert Blendon, find evidence of a disappointing "group
think" in the questions, particularly in recent elections. Andy Kohut agrees that
the questionnaire has become "too homogenized" because of the group effort,
and suggests the NEP hire someone to design the best questionnaire possible.
Paul Lavrakas thinks the NEP needs a strong academic voice guiding
questionnaire development. Jon Krosnick, who has done extensive research on
survey questionnaires, also worries about the "sub-optimalities" of the current
exit poll questionnaire design, and urges the NEP to do further study on
improving questionnaire quality.


Since the answers to the exit poll questions provide reporters in the first post-
election news cycle the single most important editorial information as they
frame the election mandate, getting the questions right is all-important.
Looking back at the 2004 exit poll, the question about "moral values" is still a
highly controversial one. The exit poll found that "moral values" topped the
list of issues deemed most important to voters. Jon Krosnick worries that the
question forced people to rationalize their vote choice, while giving them
limited options to choose among. The National Journal covered the issue:


                                         24
       Designers of this year's most widely used exit poll defend
       inclusion of the "moral values" category, which was chosen by 22
       percent of voters, because the Bush re-election campaign focused
       on the phrase. But to the poll's critics, "moral values" is a vague,
       appealing catchall category that distorts more than it reveals about
       voters' political attitudes. One objecting pollster said that adding
       "moral values" to a list of more conventional issue options was
       much like asking voters, "`what do you like best--red, green, blue,
       or breathing?"49

The National Journal points out the "real-world political consequences" of
misinterpreting election results. In 2004, groups opposed to abortion rights
and rights for gays cited the exit poll "moral values" mandate as evidence that
they deserved legislative rewards from the president and the Republican-
controlled Congress. Richard Morin of the Washington Post was harsh in his
criticism, saying the "moral values" issue added to the other problems of the
2004 exit poll:


       It seems clear now that the 2004 exit polls were rife with
       problems, most of them small but none trivial. Skewed samples,
       technical glitches and a woefully inept question that included the
       undefined term "moral values," in a list of concrete issues all
       combined to give exit polling its third black eye in as many
       elections.50

Robert Blendon and Michael Traugott, both deeply appreciative of the rich
data about the electorate provided by exit polls in the last 40 years, worry about
the future of the exit poll analytical data. All of the polling experts consulted
thought that in the pre-1990 era of exit polling, when the news organizations
each did their own polling, the product was more robust and provided a better
collective judgment about the election results.



                                        25
Projections
After the 2000 election, ABC News announced it would project the winner in a
race in a given state only after the last scheduled poll closing time in that state.51
The other NEP members made similar announcements. Prior to that time,
races were sometimes called when the "overwhelming majority" of polls in a
given state were closed, which is why the initial 2000 Florida projection was
made when polls were still open in the western panhandle. Many politicians
were hoping for more restraint, again raising the projection issues first
discussed with Congress in the 1980s. Since the 1980s, however, the birth of
cable TV and the explosion of news sites on the internet have created a high-
speed 24-hour news cycle. Leaked exit poll data, as discussed before, have been
widely reported in the blogosphere by "citizen journalists" bound by no
agreements with the NEP, or Congress for that matter, about guidelines for
reporting on elections. No longer is the issue of discussing election reporting
in the hands of three gentlemen network news division presidents, as it was in
the 1980s.


Nevertheless, in 2005's "Report on the Commission on Federal Election
Reform," the following recommendation was made by chairs Jimmy Carter and
James Baker: "News organizations should voluntarily refrain from projecting
any presidential election results in any state until all of the polls have closed in
the 48 contiguous states."52 Instead, news organizations continue to counter
that a uniform poll closing time across the country would resolve the issue,
since at least the NEP members have voluntarily agreed not to project races
until each state's polls are closed.




                                         26
In reality, since the 2000 election, news organizations are reluctant to be first
calling the closest elections. The ultra-competitive race to be first slowed down
after the embarrassment of calling Florida incorrectly in 2000. Exit polls had
been so accurate up to that point that the level of confidence in the data was
high. In the reviews of their 2000 coverage, the networks each discussed
internal changes to be made in the way they would call races in the future,
including how and where they would run their decision desks, terminology they
would use in estimating winners, and circumstances under which they would
not rely on exit poll data to project winners.53


Conclusion
Exit polling was the right methodology at the right time when it was developed
in the 1960s. No academic or polling expert would invent the same system
today. It needs to be reinvented or replaced after the 2008 election.


With the changing ways Americans are voting, the many problems associated
with all types of polling, the ability to provide deep and instant data to
individual news consumers on the internet, and the ability to use the internet to
reach individual voters, there are challenges and opportunities for new ways to
measure election results that did not exist in the 1960s.


The challenges include finding new ways to survey a large sample of
representative American voters, whether they vote early or on Election Day, in
order to tell a timely and accurate story about election results. A concomitant
challenge for the news organizations involved is finding ways to financially
support these efforts. It is a very expensive proposition to undertake what Paul




                                        27
Lavrakas calls "the single largest data collection and analysis that takes place in
one day on the planet."


There are many suggestions about how to do this post-2008. A particularly
difficult issue is how to achieve the same individualized, 50-state data collection
using any other methodology. Joe Lenski, Paul Lavrakas, and others would like
to see the NEP invest in research and development to overhaul the exit poll
methodology, developing new statistical sampling models. New models would
have to address today's cycle-to-cycle changes in early voting patterns and
precinct demographics. Opinion Dynamics chairman John Gorman thinks a
re-do should start from scratch, that there is "too much modeling and not
enough straight-forward polling" in the current system, and that it is too
complicated.54 Several survey research experts suggested small changes, such as
modernizing information delivery from the field, upgrading from paper ballots
to using wireless handheld devices that instantly transmit results.


In 2006, when he was asked about the future of exit polling, Warren Mitofsky
said he expected projection models would have to change.55 Others have
suggested replacing the current exit poll with a hybrid of "best of"
methodologies: a smaller national exit poll in key states coupled with rich pre-
election telephone polls, Election Day telephone polls, and an internet panel.
In Norway, the use of a "telepanel" as an Election Day poll has been quite
successful, and was watched with interest by Warren Mitofsky. A panel is
recruited and interviewed by telephone a short time ahead of the election, and
then re-interviewed on Election Day.56 This is a methodology considerably
cheaper than exit polling, although it would be difficult to accomplish a 50-
state breakdown of results.


                                        28
Many academics think internet polling is the future, particularly with 71% of
American adults now using the internet, and all other polling methodologies
under siege.57 No one suggests a total reliance on internet polling at this point,
though, because of the sampling limitations. Minorities and senior citizens are
still under-represented among internet users. Many internet polls are "opt-in"
and considered unreliable and unscientific. However, some academics and
news organizations are already experimenting with using different types of
internet panels to question voters. Steve Ansolabehere of MIT had success with
a large internet panel in 2006, using a matched random sample weighted to
compensate for under-representation of certain demographic groups, and will
do the same in 2008.58 The Associated Press-Yahoo is working with
Knowledge Networks to survey voters throughout 2008 using a probability-
based online panel designed to represent the U.S. population. Jon Krosnick of
Stanford has a grant from the National Science Foundation to experiment with
"best of" polling methodologies. These efforts deserve careful review.


The opportunities include using the internet in new ways: providing transparent
and searchable exit poll data to individual users; delivering real-time election
results directly to users; explaining and archiving poll results for future
research. In some ways, the rich data provided by exit polls could be even
more important in the internet age. The many thousands of people who
searched the internet for leaked exit poll data in 2004 showed that there is a real
market for even faster delivery of incremental election reporting. The exit poll
data, once the purview of experts and academics, could be more accessible to
all. When he was with Hotline, NBC political director Chuck Todd proposed a
for-profit model of exit polling as a new business. He thought there would be


                                         29
an audience for a pay-per-view TV show on Election Day, with every wave of
exit poll information shown directly to viewers. Reporters would explain all the
shortcomings of the early waves of data, but viewers would have the same
access to live information that news organizations now have. In a 2004
election Online Journalism Review post, Slate media critic Jack Shafer had a similar
idea about delivering and demystifying exit poll results:


       The exit poll numbers are being swapped from NEP to its clients
       to politicians and journalists to boardroom big shots today like
       crazy, so why shouldn't civilians have access to the information? I
       trust readers and viewers to see the exit polls for what they are.59

The NEP will conduct exit polls again for the 2008 election, making every
effort to compensate for the problems and challenges already discussed,
spending millions of dollars to gather the data. In the event of another very
close presidential election, it is quite likely that exit poll results will again be
controversial. But whether exit polling is replaced or reinvented after 2008,
news organizations will still rush to tell voters who won an election and why. It
is what American voters want to know, and deserve to know, on Election Day.




                                           30
Acknowledgments

Special thanks to ABC News president David Westin and Senior Vice President
Kerry Smith for their support for this project, and for allowing me the time
away from ABC to work on it. Thanks also to the Shorenstein Center staff,
particularly Edie Holway, Nancy Palmer, and Tom Patterson for their help and
advice, and to the other Shorenstein Fellows, Tom Fiedler, Geoff Cowan and
Tuan Nguyen, for their enthusiastic support. Professors Robert Blendon and
Marion Just provided invaluable insight and guidance, and their help was
greatly appreciated. Finally, I want to acknowledge the terrific work done by
my research assistant Carlyn Reichel. Her contributions were many.




                                     31
Interviews

Kerry Smith, ABC News
Dan Merkle, ABC News
Gary Langer, ABC News
Cokie Roberts, ABC News
Mark Halperin, ABC News and Time magazine
Sandy Johnson, Associated Press
Maralee Schwartz, Washington Post
Amy Walter, National Journal
Chuck Todd, NBC News
Robert Blendon, Harvard University
Marion Just, Wellesley College and Harvard University
John Della Volpe, Harvard University
John Palfrey, Harvard University
Andy Kohut, PEW Research Center
Steve Ansolabehere, MIT
Paul Lavrakas, Survey Methodologist
Michael Traugott, University of Michigan
Jon Krosnick, Stanford University
John Mark Hansen, University of Chicago
Joe Lenski, Edison/Mitofsky
John Gorman, Opinion Dynamics
Mark Blumenthal, Pollster.com




                                     32
Sources

Traugott, Michael W. and Paul J. Lavrakas, The Voter's Guide to Election Polls.
Lanham: Rowman & Littlefield Publishers, Inc., 2004.

Lavrakas, Paul J. & Michael Traugott (Eds.), Election Polls, the News Media, and
Democracy. New York: Chatham House, 2000.

Fortier, John C., Absentee and Early Voting; Trends, Promises, and Perils,
Washington, D.C.: The AEI Press, 2006.

Crigler, Ann N., Marion R. Just, Edward J. McCaffery, Rethinking the Vote, New
York: Oxford University Press, 2004.

Genovese, Michael A., and Matthews J. Straub (Eds.) Polls and Politics, The
Dilemmas of Democracy. New York: State University of New York Press, 2004.

Freeman, Steven F., and Joel Bleifuss, Was the 2004 Presidential Election Stolen?
Exit Polls, Election Fraud, and the Official Count, New York: Seven Stories Press,
2006.

Stimson, James A., Tides of Consent, How Public Opinion Shapes American Politics,
Cambridge, UK: Cambridge University Press, 2004.

Blumenthal, Mark. "Toward an Open-Source Methodology, What We can
Learn from the Blogosphere." Public Opinion Quarterly, Vol. 69, No. 5,
Special Issue, 2005, pp 655-669.

Plissner, Martin. "In Place of Many, One." Public Opinion Pros, 2004.
www.publicopinionpros.com/features/2004/nov/plissner_1htm

Langer, Gary, and Cohen, Jon. "Voters and Values in the 2004 Election."
Public Opinion Quarterly, Vol. 69, No. 5, Special Issue, 2005, pp. 744-759.

Richman, Alvin. "Gauging the Election Mandate and the Role of Moral
Values." Public Opinion Pros, 2005.
www.publicopinionpros.com/features/2005/mar/richman.asp

Kennedy, Robert F., Jr. "Was the 2004 Election Stolen?" Posted June 1, 2006.
www.rollingstone.com/news/story/10432334/was_the_2004_election_stolen


                                          33
Liddle, Elizabeth. "Edison/Mitofsky Exit Polls 2004: Differential Non-
Response or Vote-Count Corruption." Posted April 27, 2005.
www.geocities.com/lizzielid/WPEpaperARCH.pdf

Election Science Institute. "Ohio 2004 Exit Polls: Explaining the Discrepancy."
Paper: June 6, 2005.

Edison Media Research and Mitofsky International. "Evaluation of E/M
Election System 2004." Paper prepared for the National Election Pool, NEP,
January 19, 2005.

Lindeman, Mark. "Condemned to Repetition: The 2006 Exit Poll
Controversy." Public Opinion Pros, 2007.
www.publicopinionpros.com/features/2007/jan/lindeman/printable.asp

Traugott, Santa. "Validating Self-Reported Vote: 1964-1988." Paper prepared
for the annual meeting of the American Statistical Association, Washington,
D.C., August 7-10, 1989.

Berinsky, Adam. "The Perverse Consequences of Electoral Reform in the
United States." American Politics Research, Vol. 33, No. 4, July 2005, pp. 471-
491.

Hammer, Michael J., and Traugott, Michael W. "The Impact of Voting by Mail
on Voter Behavior." American Politics Research, Vol. 32, No. 4, July 2004, pp.
355-405.

Stein, Robert. "Early Voting." Public Opinion Research, Vol. 62, 1998, pp.
57-69.

Mason, Linda and Frankovic, Kathleen, and Jamieson, Kathleen Hall. "CBS
News Coverage of Election night 2000." Paper prepared for CBS News,
January 2001.

Goldstein, Tom and McCormick, David. "Summary of Election Night
Review." Paper prepared for NBC News, January 4, 2001.

Westin, David. "Election Night 2000 Coverage by the Networks." Testimony
before the Full Committee on Energy and Commerce, U.S. House of
Representatives, delivered on February 14, 2001.


                                      34
Mitofsky, Warren. "Voter News Service After the Fall," Public Opinion
Quarterly. Vol. 62, 2003, pp. 45-58.




                                    35
Notes

1
 Lawrence K. Grossman, "Exit Polls, Academy Awards, and Presidential Elections..."
Columbia Journalism Review, May 1, 2000.
2
 Mark Lindeman and Rick Brady, "Behind the Controversy: A Primer on U.S. Exit Polls,"
Public Opinion Pros, 2006.
http://www.publicopinionpros.com/from_field/2006/jan/lindeman_1.asp
3
 Joan Konner, James Risser, and Ben Wattenberg, "Television's Performance on Election
Night 2000, A Report for CNN." January 29, 2001, pp. 32-33.
http://election2000.stanford.edu/cnnelectionreport.pdf
4
 Information summarized from reports on the 2000 election prepared by ABC, CBS, CNN,
and NBC. See Source listings.
5
 Final Report of the National Commission on Election Reform. "To Assure Pride and
Confidence in the Electoral Process," August 2001, p. 64.
http://www.tcf.org/Publications/ElectionReform/99_full_report.pdf
6
 Michael Traugott, "Recovering from Disaster: The Path Ahead," International
Communication Association & American Political Science Association, Vol. 13, No. 2,
Spring 2003.
7
    Joan Konner, "The Case for Caution," Public Opinion Quarterly, Volume 67, 2003, pp. 5-18.
8
 Curtis Gans, "Making It Easier Doesn't Work--No Excuse Absentee and Early Voting
Hurt Voter Turnout; Create Other Problems," Committee for the Study of the American
Electorate, press release, September 13, 2004; Paul Gronke, Eva Galanes-Rosenbaum, Peter
A. Miller, "Early Voting and Turnout," PSOnline, www.apsanet.org, October 2007.
9
 Early voting data source: Information provided by Edison/Mitofsky to NEP members in
2006.
10
  Robert M. Stein, Jan Leighley, Christopher Owens, "Who votes, who doesn't, why and,
what can be done?" A Report to the Federal Commission on Electoral Reform, June 10,
2005.
http://www.american.edu/ia/cfer/0630test/stein.pdf
11
  Data from California's Secretary of State website, "Historical Absentee Ballot Use in
California," accessed December 11, 2007.
http://www.sos.ca.gov/elections/hist_absentee.htm
12
  Early voting data source: Information provided by Edison/Mitofsky to NEP members in
2006.


                                               36
13
     Paul Lavrakas, interviewed by telephone, November 13, 2007.
14
     Joe Lenski, interviewed by telephone, December 3, 2007.
15
     Marion Just, interviewed in person on numerous occasions, September­December 2007.
16
  Paul Gronke, "Early Voting Reforms and American Elections," Paper presented at the
Annual Meeting of the American Political Science Association, Chicago, IL, September 2-5,
2004, p. 2.
17
     Ibid., pp. 9-10.
18
  Matt A. Barreto, Matthew J. Streb, Mara Marks, Fernando Guerra, "Do Absentee Voters
Differ From Polling Place Voters?" Public Opinion Quarterly, Vol. 70, No. 2, Summer 2006,
pp. 224-225.
19
  Merkle, D.M. and Edelman, M., "A Review of the 1996 Voter News Service Exit Polls
from a Total Survey Error Perspective." In Paul.J. Lavrakas & Michael Traugott (Eds.),
Election Polls, the News Media, and Democracy, New York: Chatham House, 2000, p. 81.
20
   Thomas Beaumont, "Dems outdo GOP in early voting," DesMoinesRegister.com,
September 28, 2006.
http://www.desmoinesregister.com/apps/pbcs.dll/article?AID=20060
21
  "Evaluation of Edison/Mitofsky Election System 2004" prepared by Edison Media
Research and Mitofsky International for the National Election Pool (NEP), January 19,
2005, p. 74.
22
     Konner, Risen, and Wattenberg, p. 25.
23
  Early voting data source: Information provided by Edison/Mitofsky to NEP members in
2006.
24
 Gerald C. Wright, "Are Exit Polls Bad for Democracy?" In Michael A. Genovese and
Matthews J. Straub (Eds.) Polls and Politics, The Dilemmas of Democracy. New York: State
University of New York Press, 2004, p. 121.
25
  Stephen J. Blumberg and Julian V. Luke, "Wireless Substitution: Early Release of
Estimates Based on Data from the National Health Interview Survey, July-December 2006,"
CDC, released May 14, 2007.
26
  Press Release: "Breaking News: Census Figures Confirm 2006 Jump in Young Voter
Turnout," Young Voter Strategies, June 15, 2007.
http://www.youngvoterstrategies.org/index.php?tg=articles&topics=2&mylocation=News


                                              37
27
  Mark Hugo Lopez, Karlo Barrios Marcelo, and Emily Hoban Kirby, "Youth Voter
Turnout Increases in 2006," CIRCLE Fact Sheet, June 2007.
http://www.civicyouth.org/PopUps/FactSheets/FS07_2006MidtermCPS.pdf
28
 David C. King, "Youth Came Through with Big Turnout," Boston Globe Op-Ed,
November 4, 2004, p. A-15.
29
  Scott Keeter, "The Impact of Cell Phone Noncoverage Bias on Polling in the 2004
Presidential Election," Public Opinion Quarterly, Vol. 70, No. 1, Spring 2006, p. 88.
30
     Konner, p.13 ­ Figures for 1992 and 1996.
31
     Figures for 2000, 2004, and 2006 provided by Edison/Mitofsky.
32
  Warren Mitofsky, "The Future of Exit Polling," Public Opinion Pros, 2006.
http://www.publicopinionpros.com/op_ed/2006/jan/mitofsky.asp
33
  Project for Excellence in Journalism, "Public Attitudes," in The State of the News Media
2007, c. 2006.
http://www.stateofthenewsmedia.org/2007/narrative_overview_publicattitudes.asp?cat=8&
media=1
34
  Merkle, Daniel M., and Edelman, Murray. "Nonresponse in Exit Polls: A Comprehensive
Analysis." In Survey Nonresponse, ed. R.M. Groves, D.A. Dillman, J.L. Eltinge, and R.J.A.
Little. New York: Wiley, 2002, pp. 243-258.
35
     Lindeman and Brady.
36
     2004 and 2006 data on differential response provided by E/M to NEP members.
37
     Will Lester, "Better Interviews Said Key to Exit Polls," Associated Press, May 15, 2005.
38
     Traugott, 2003.
39
     Jon Krosnick, interviewed by telephone, October 18, 2007.
40
  Michael Traugott, Benjamin Highton, Henry E, Brady, "A Review of Recent
Controversies Concerning the 2004 Presidential Election Polls," for The National Research
Commission on Elections and Voting, March 10, 2005, p. 8.
41
     Andy Kohut, interviewed by telephone, September 28, 2000.
42
  Richard Morin, "Surveying the Damage; Exit Polls Can't Predict Winners, So Don't
Expect Them To," Washington Post, November 21, 2004, p. B01.


                                               38
43
  Paul Lavrakas, "Why Our Democracy Needs Accurate National Exit Polls," Public Opinion
Pros.
http://www.publicopinionpros.com/features/2007/jan/lavrakas_printable.asp
43
     Mark Blumenthal, interviewed by telephone, November 13, 2007.
45
     Data provided to NEP members by Edison/Mitofsky.
46
     John F. Harris, Washington Post, December 26, 2004, p. A06.
47
  Steven F. Freeman., Ken Warren, Stephanie Frank Singer, "The Need for Independent US
Exit Polls," PFAW Proposal--Election Verification Exit Poll, July 4, 2006.
48
  Linda Mason, Kathleen Frankovic, Kathleen Hall Jamieson, "CBS News Coverage of
Election Night 2000" CBS News, 2001, p.68.
49
     Brian Friel, "Polling's Moral Dilemma," National Journal, November 20, 2004.
50
  Rich