The Significance of a Popular Vote Victory

Although it is the Electoral College and not the popular vote that legally elects the
president, winning the popular vote does have considerable psychological and practical
significance. It is fair to say, to take a recent example, that had Al Gore not enjoyed a
popular vote margin in 2000, he would not have had standing in the court of public
opinion to maintain his post-election challenge for more than a month up until its
ultimate foreclosure by the Supreme Court.

In the 2004 election now under scrutiny, the popular vote again has played a
critical role. George Bush's apparent margin of 3.3 million votes clearly influenced the
timing of John Kerry's concession. Although the election was once again close enough
that yet-to-be-counted votes offered at least the mathematical possibility of a Kerry
electoral college victory--and although, once again, concerns about vote counting were
beginning to emerge from early post-election reports and analyses--Kerry apparently
believed that, unlike popular vote-winner Gore, he did not have effective standing to
prolong the race.

As ongoing inquiries continue to raise serious vote counting issues, Bush's
apparent popular vote margin has loomed large as a rationale for minimizing these issues,
at least as far as their impact on the outcome of the race. While much concern has been
expressed about "counting every vote," even the Kerry camp has issued disclaimers to
the effect that their candidate does not expect that so doing will alter the outcome.

With the results in Ohio currently subject to both recount proceedings and legal
contest, dramatic developments compelling a reversal of the Ohio result cannot be ruled
out at this time. Yet to overturn the Ohio result, giving Kerry an electoral college victory
(or even to disqualify the Ohio electors via challenge in Congress, which would deprive
Bush of an electoral college majority and throw the election to the House of
Representatives), would likely be regarded as unjust and insupportable by a populace
convinced that Bush was, by some 3.3 million votes, the people's choice.

Thus, although the popular vote does not legally determine the presidency, its
significance is such that we must give due consideration to any evidence which puts the
popular vote count itself at issue.

Sources of the Exit Poll and Vote Count Numbers

As the analysis which we undertake below is based upon the conflict between two
sets of numbers, one generated by the exit polls for the presidential race and the other
generated by the vote counting equipment, it is necessary to review the nature of the two
sources of results. Exit polling, since its invention several decades ago, has performed
reliably in the projection of thousands of races, both here at home and, more recently,
abroad.1 The record of exit polling from the 1970s through the 1990s was essentially free
of controversy, except for the complaint that publication of exit poll results prior to poll
closings dampened voter turnout by discouraging late-in-day voters from bothering to
vote, the race having already been "called."2 Voters could be so influenced because they
had come, indeed, to regard exit poll projections as all but infallible. Significant exit
polling problems began to appear along with the development and spread of
computerized vote counting equipment, since which time exit polls have had a notably
poorer track record in spite of improvements in polling methodology.

Compared to standard pre-election polling, exit polling has certain advantages
and disadvantages. On the plus side, exit polls sample actual rather than just "likely"
voters and do not fail to include voters who are not attached to a conventional phone line
or who screen their calls.3 This results in significantly greater accuracy. On the minus
side, exit polls employ a cluster sampling technique, grouping respondents by precinct,
rather than a fully homogenized random sample of the target venue. This results in
somewhat less accuracy. On the whole, the advantages in accuracy an exit poll enjoys
over a pre-election poll of the same sample size tend to outweigh the disadvantages.
The exit polling in Election 2004 was performed by the combined firms of
Mitofsky International and Edison Media Research, under exclusive contract as "official
provider" of exit poll data to six major media organizations (CBS, NBC, ABC, CNN, Fox
News Channel, and the Associated Press), which collectively formed the National
Election Pool. 4 Exit polling operations were under the principal direction of Warren
Mitofsky, credited as the inventor of exit polling and recognized throughout the world as
the leading expert in the field. With over 35 years of exit polling experience,
encompassing nearly 3,000 electoral contests in the United States and abroad, Mitofsky
has achieved consistent success in the field and has continued throughout his career to
refine and improve the methodologies and protocols of exit polling.5 In 1999 Mitofsky
received the Award for Lifetime Achievement from the American Association for Public
Opinion Research.

Election 2004 presented a particular challenge and opportunity for Mr. Mitofsky,
whose exit polling operation was hampered in 2002 by a massive computer breakdown.6
It has been reported that preparations for Election 2004 were especially thorough,
entailing increased staff numbers and training, upgraded computer hardware and
software, expanded surveys of absentee and early voters, and dry runs beginning in July
to prepare analysts for the full spectrum of possible election night scenarios.7 It may
fairly be said that the exit polling for Election 2004 was a more advanced, sophisticated,
and meticulous operation than any previously undertaken.

In contrast to the uniform methodology of the exit polls, a variety of methods are
employed to record votes on election day, including optical scan devices, direct
electronic recording (DREs or "touchscreens"), punch cards, paper ballots, lever
machines, and data-point devices, in that order of prevalence. An additional variety of
methods are then employed to transmit these votes to central locations and tally them at
the county and state levels. Ownership and operation of this mosaic of machinery is fully
privatized and is concentrated predominantly in the hands of four corporations: Diebold,
ES&S, Sequoia, and Hart Intercivic. The partisan proclivities and activities of each of
these corporations are a matter of public record. 8 Because of the proprietary nature of
the election system throughout the United States, these vendors of the voting equipment
design, program, operate, maintain, and repair it at every level, most often without
outside or public scrutiny, and with at best a minimal process of testing and certification.9
Boards of Election and state level authorities over election protocols have often accepted
financial support from the equipment vendors10 and have also been seen at times to act
under the influence of partisanship, appearing to elevate outcomes over fairness of
process.11 Such systemic conflicts of interest do little to enhance the integrity or
credibility of the vote counting system.

Computer experts have documented the susceptibility of both the recording and
tabulating equipment to undetected errors, hacking, and deliberate fraud.12 A substantial
component of the system (DREs, which are responsible for recording approximately 30%
of the vote) generates no paper record and is effectively immune to meaningful recount.
Central tabulators responsible for compiling over 50% of the vote employ an operating
system that has been demonstrated to be vulnerable to entry and manipulation through a
standard laptop PC.13 In spite of these vulnerabilities of the counting system, few if any
questions about the accuracy of the numbers it produced were raised on election night.14

Election Night 2004: The Exit Poll/Vote Count Differential

On election night 2004, the exit polls and the vote counting equipment generated
results that differed significantly. In the early morning of November 3, 2004, a
CNN.com website screenshot entitled "U.S. PRESIDENT/NATIONAL/EXIT POLL"
posted national exit poll results updated to 12:23 A.M., broken down by gender as well as
a variety of other categories.15 The time of the update indicates that these results
comprised substantially the full set of respondents polled on election day, but were free
from the effects of a subsequent input of tabulated data used to bring about ultimate
congruence between the exit poll and vote count results.16

The CNN posting indicates the number of respondents (13,047), the gender
breakdown of the sample (male 46%, female 54%), and the candidate preferences by
gender (males: 52% Bush, 47% Kerry; females: 45% Bush, 54% Kerry). For the national
exit poll taken as a whole, therefore, the result was 48.2% Bush, 50.8% Kerry.17 The vote
counting equipment produced a markedly different result: 50.9% Bush, 48.1% Kerry.18

The differential between the two counts, which were virtually mirror images of each
other, was 5.4% overall, or about 3.3 million votes (see Chart 1).

The reaction of election night analysts interpreting this differential was
immediately to query what had "gone wrong" with the exit polls. This was a curious
approach both in light of standard accounting practice, which compels independent
examination of both sets of numbers that are found to be in conflict, and in light of muchvoiced
pre-election concerns about the accuracy and security of the computerized vote
counting systems. We offer an alternate approach to the conflicting data, based on
fundamental statistical and accounting principles.

Statistical Analysis Of Exit Poll Results

Steven F. Freeman of the University of Pennsylvania has analyzed Election 2004
exit poll results for battleground states,19 and has drawn certain conclusions regarding the
significant discrepancies between exit poll results and vote counts for several critical
states. In particular, the odds against the discrepancies in Ohio, Florida, and
Pennsylvania occurring together are computed at 662,000-to-one, or a virtual statistical
impossibility that they could have been due to chance or random error. Receiving
somewhat less emphasis is the overall pattern of discrepancy in the state polls--again
with the vote counts turning in Bush's favor, though less dramatically in the
nonbattleground states, as will be discussed below. The national popular vote is not
addressed in that paper, but the same statistical principles are applicable, and will be
employed in this analysis.

While the individual state samples totaled 73,678 reported respondents,20 a
national sub-sampling was undertaken by Edison/Mitofsky, which comprised 13,047
reported respondents, chosen as a representative random sample of the nation as a whole.
This sample was drawn from 250 targeted polling places and from 500 individual
telephone interviews with absentee and early voters.21

What is remarkable about this national sample of 13,047 is its size. When
compared with more familiar pre-election poll samples of about 2000 - 2200 respondents,
it is approximately six times as large. Such augmentation of sample size reduces a poll's
margin of error (MOE) from the ±3% to which we have become accustomed, down to
±1.1%.22
The ±1.1% MOE tells us that, barring specific flaws in the design or
administration of the poll and in the absence of significant mistabulation of the vote
count itself, the exit poll result for the selected candidate will fall within ±1.1% of his
vote count 95% of the time. In this case it tells us that we can be 95% certain that
Kerry's popular vote percentage would fall in the range 49.7% to 51.9%; that is, it would
fall outside that range only once in 20 times. Kerry's reported vote count of 48.1% falls
dramatically outside this range.23

To carry our analysis further, we can employ a normal distribution curve (see
Figure 1) to determine--again assuming proper poll methodology, no discriminatory
voter suppression 24, and an accurate and honest popular vote count--that the probability
that Kerry would have received his reported popular vote total of 48.1%, or less is one in
959,000--a virtual statistical impossibility.25

The Popular Vote Winner

We can proceed one helpful step further and calculate the likelihood, based on the
exit poll results, that Kerry would receive more popular votes than Bush. The break-even
point would be 59,024,629 votes, or 49.54% of the total.26 This percentage lies,
significantly, outside the MOE of the national exit poll and in fact we find that Kerry
would receive fewer votes than Bush only 1.3% of the time. Put another way--given the
exit poll results, proper poll methodology, and an accurate and fair voting process--
Kerry would be the popular vote winner of Election 2004 98.7% of the time.

Is Something Wrong With The Exit Poll Results?

The clear implication of our analysis is that neither chance nor random error is
responsible for the significant incongruence of exit poll and tabulated vote results, and
that we must look either to significant failings in the exit poll design and/or
administration or to
equally significant failings in the accuracy and/or fairness of the voting process itself to
explain the results. Given the dramatic implications of our analysis, we of course must
consider carefully any argument that has been put forward suggesting that the exit polls
failed as an accurate measure of voter intent. We examine the two least implausible
hypotheses that have been put forward.

The first deals with the proportion of respondents by gender. The composition of
the national sample by gender was 46% male, 54% female, which prompted a claim that
females were over-represented, skewing the results towards Kerry. While it is not proven
that this is in fact the case, if it is taken as stipulated and the sample is reweighted to
reflect a "normal" gender breakdown of 52% female, 48% male, the effect is to increase
Bush's exit poll percentage by 0.2% to 48.4% and decrease Kerry's to 50.6%. The effect
on the bottom line is minimal: Kerry would be the popular vote victor 96.9% of the
time.27

The second hypothesis put forward is the "reluctant Bush responder" hypothesis.
It suggests that Bush voters were for some reason less willing to fill out an exit poll
questionnaire, and therefore were undercounted in the poll results. If such a phenomenon
could be proven, it would be a source of significant skewing and effectively invalidate
the polls. The proponents of this hypothesis, however, have yet to offer any supportive
evidence for their theory.28 The hypothesis also does not explain the nonuniformity of
the pattern of state-by-state discrepancies.29 In fact, one could equally well imagine that
a "reluctant Kerry responder" phenomenon was at work, and that the exit polls
systematically underrepresented Kerry's vote.30

Conclusion

In light of the history of exit polling and the particular care that was taken to
achieve an unprecedented degree of accuracy in the exit polls for Election 2004, there is
little to suggest significant flaws in the design or administration of the official exit polls.

Until supportive evidence can be presented for any hypothesis to the contrary, it must be
concluded that the exit polls, including the national mega-sample within its ±1.1%
margin of error, present us with an accurate measure of the intent of the voters in the
presidential election of 2004.

According to this measure, an honest and fair voting process would have been
more likely than not--at least 95% likely, in fact--to have determined John Kerry to be
the national popular vote winner of Election 2004.31 Should ongoing or new
investigations continue to produce evidence that, to an extent determinative of the
electoral college outcome, votes have not been counted accurately and honestly or
discriminatory vote suppression has occurred, the re-examined popular vote outcome
may well be deemed relevant to the question of what remedies are warranted.