| (page two)
In Dissent
A response to the Florida election report.
By Commissioners Abigail
Thernstrom & Russell G. Redenbaugh
II. The Evidence Fails To Support
the Claim of Systematic Disfranchisement
Based on witnesses' limited (and often, uncorroborated)
accounts, the Commission insists that there were
"countless allegations" involving "countless
numbers" of Floridians who were denied the right to vote.
This anecdotal evidence is drawn from the testimony of 26
"fact witnesses," residing in only eight of the state's 67
counties.
In fact, however, many of those who appeared before the
Commission testified to the absence of "systemic
disenfranchisement" in Florida. Thus, a representative of the
League of Women Voters testified that there had been many
administrative problems, but stated: "We don't have any
evidence of race-based problems
we actually I guess don't
have any evidence of partisan problems." And a witness from
Miami-Dade County, who said she attributed the problems
she encountered not to race but rather to inefficient poll
workers: "I think [there are] a lot of people that are on jobs
that really don't fit them or they are not fit to be in."
Without question, some voters did encounter difficulties at the
polls, but the evidence fails to support the claim of systematic
disfranchisement. Most of the complaints the Commission
heard in direct testimony involved individuals who arrived at
the polls on election day only to find that their names were
not on the rolls of registered voters. The majority of these
cases many due to bureaucratic errors, inefficiencies within
the system, and/or error or confusion on the part of the voters
themselves.
III The Commission Failed to Distinguish Between Bureaucratic
Problems and Actual Discrimination
Other witnesses did offer testimony suggesting numerous
problems on election day. But the Commission, in discussing
these problems, failed to distinguish between mere
inconvenience, difficulties caused by bureaucratic
inefficiencies, and incidents of potential discrimination. In its
report, the complaint from the voter whose shoes were
muddied on the path to his polling place is accorded the same
degree of seriousness as the case of the seeing-impaired
voter who required help in reading the ballot, or the African
American voter who claimed she was turned away from the
polls at closing time while a white man was not.
There were certainly jammed phone lines, confusion and
error, but none of it added up to widespread discrimination.
Many of the difficulties, like those associated with the
"butterfly ballot," were the product of good intentions gone
awry or the presence of many first-time voters. The most
compelling testimony came from disabled voters who faced a
range of problems, including insufficient parking and
inadequate provision for wheelchair access. This problem, of
course, had no racial dimension at all.
| IV. The Majority Report's Relies Upon a Warped interpretation
of the Voting Rights Act |
The report essentially concludes that election procedures in
Florida were in violation of the Voting Rights Act, but the
Commission found no evidence to reach that conclusion
only a court can and has bent the 1965 statute totally out
of shape.
The question of a Section 2 violation can only be settled in a
federal court. Plaintiffs who charge discrimination must prevail
in a trial in which the state has a full opportunity to challenge
the evidence. To prevail, plaintiffs must show that "racial
politics dominate the electoral process," as the 1982 Senate
Judiciary Committee Report stated in explaining the newly
amended Section 2.
The majority's report implies that Section 2 aimed to correct
all possible inequalities in the electoral process. Had that been
the goal, racially disparate registration and turnout rates
found nearly everywhere in the country would constitute a
Voting Rights Act violation. Less affluent, less educated
citizens tend to register and vote at lower rates, and, for the
same reasons, are likely to make more errors in casting
ballots, especially if they are first time voters. Neither the
failure to register nor the failure to cast a ballot properly as
regrettable as they are are Section 2 violations.
Thus, despite the thousands of voting rights cases on the
books, the majority report cannot cite any case law that
suggests punch card ballots, for instance, are potentially
discriminatory. Or that higher error rates among black voters
suggest disfranchisement.
There is good reason why claims brought under Section 2
must be settled in a federal court. The provision requires the
adjudication of competing claims about equal electoral
opportunity an inquiry into the complex issue of racial
fairness. The Commission is not a court and cannot arrive at
verdicts that belong exclusively to the judiciary. Yet, while the
majority report does admit that the Commission cannot
determine if violations of the Voting Rights Act have actually
occurred, in fact it unequivocally claims to have found
"disenfranchisement," under the terms of the statute.
| V. Misplaced Responsibility for Election Procedures |
The report holds Florida's public officials, particularly the
governor and secretary of state responsible for the
discrimination that it alleges. "State officials failed to fulfill their
duties in a manner that would prevent this
disenfranchisement," it asserts. In fact, most of the authority
over elections in Florida resides with officials in the state's 67
counties, and all of those with the highest rates of voter error
were under Democratic control.
The report charges that the governor, the secretary of state
and other state officials should have acted differently in
anticipation of the high turnout of voters. What the
Commission actually heard from "key officials" and experts
was that the increase in registration, on average, was no
different than in previous years; that since the development of
"motor voter" registration, voter registration is more of an
ongoing process and does not reach the intensity it used to
just prior to an election; and that, in any event, registration is
not always a reliable predictor for turnout.
There was a 65 percent increase in African American voters,
40 percent of whom were coming to the polls for the first
time. But this was an unanticipated event.
The majority report also faults Florida state officials with
having failed to provide the 67 supervisors of elections with
"adequate guidance or funding" for voter education and
training of election officials. What the report pointedly ignores
is that the county supervisors are independent, constitutional
officers who make their budget requests to the boards of
county commissioners, not to the state.
| VI. The Commission Provides Only A One-Sided Examination of the
Felon List |
The report asserts that the use of a convicted felons list "has a
disparate impact on African Americans." "African Americans
in Florida were more likely to find their names on the list than
persons of other races." Of course, because a higher
proportion of blacks have been convicted of felonies in
Florida, as elsewhere in the nation. But there is no evidence
that the state targeted blacks in a discriminatory manner in
constructing a purge list, or that the state made less of an
effort to notify listed African Americans and to correct errors
than it did with whites. The Commission did not hear from a
single witness who was actually prevented from voting as a
result of being erroneously identified as a felon. Furthermore,
whites were twice as likely as blacks to be placed on the list
erroneously not the other way around.
The compilation of the purge list was part of an anti-fraud
measure enacted by the Florida legislature in the wake of a
Miami mayoral election in which ineligible voters cast ballots.
The list for the 2000 election was overinclusive, and some
supervisors made no use of it. (The majority report did not
bother to ask how many counties relied upon it.) On the other
hand, according to the Palm Beach Post, more than 6,500
ineligible felons voted.
Based on extensive research, the Miami Herald concluded
that the biggest problem with the felon list was not that it
wrongly prevented eligible voters from casting ballots, but
that it ended up allowing ineligible voters to cast a ballot. The
Commission should have looked into allegations of voter
fraud, not only with respect to ineligible felons, but allegations
involving fraudulent absentee ballots in nursing homes,
unregistered voters, and so forth. Across the country in a
variety of jurisdictions, serious questions about voter fraud
have been raised.
| VII. Unwarranted Criticism of Florida Law Enforcement |
Despite clear and direct testimony during the hearings, as well
as additional information submitted by Florida officials after
the hearings, the report continues to charge the Florida
Highway Patrol with behavior that was "perceived" by "a
number of voters" as "unusual" (and thus somehow
"intimidating") on election day. In fact, only two persons are
identified in the report as giving their reactions to activities of
the Florida Highway Patrol on election day. One testified
regarding a police checkpoint, and the other testified that he
found it "unusual" to see an empty police car parked outside
of a polling facility. Neither of these witnesses' testimony
indicates how their or others' ability to vote was impaired by
these events.
| VIII. Procedural Irregularities at the U.S. Commission on Civil
Rights |
Procedural irregularities have seriously marred the majority
report. In writing the report, the Commission ignored not only
the rules of evidence, but the agency's own procedures for
gathering evidence. By arguing that "every voice must be
heard," while in fact stifling the voice of the political minority
on the Commission itself, it is guilty of gross hypocrisy.
Among the procedural problems in the drafting of the report:
Republican-appointed commissioners were never
asked for any input in the composition of the witness
list or in the drafting of the report itself. In fact, we
were denied access to the witness lists altogether. An
outside expert, with strong partisan affiliations, was
hired to do a statistical analysis without consultation
with commissioners.
At the hearings in Florida, the secretary of state and
other Republican witnesses were treated in a manner
that fell far short of the standard of fair, equal and
courteous.
The majority reached and released its verdict, in the
form of a "preliminary assessment," long before the
report was ready for discussion.
Florida authorities who might be defamed or degraded
by the report were not given the proper time to review
the parts of the report sent to them to say nothing
of their right to review the report in its entirety.
Affected agencies were not given adequate time to
review applicable provisions, and a draft final report
was made available to the press that included no
corrections or amendments on the basis of affected
agency comments.
Commissioners were given only three days to read the
report one less day than three major newspapers
had before its approval by the Commission at the
June 8 meeting. This and other aspects of the process
were contrary to the schedule, and made careful,
detailed feedback at the time literally impossible.
In its efforts to investigate procedural irregularities in Florida,
the Commission has clearly engaged in serious procedural
irregularities of its own. By consistently violating its own
procedures for fair and objective fact-finding, the
Commission, undermines its credibility and calls into question
the validity of its work.
| Part I: The Commission's Statistical Analysis Fails to Prove
Disfranchisement |
The statistical analysis done for the Commission by Dr.
Allan Lichtman does not support the claim of
disfranchisement. The most sensational "finding" in the
majority report is the claim that black voters in Florida
were nine times as likely as other residents of the state to
have cast ballots that did not count in the presidential
contest, and that 52 percent of all disqualified ballots
were cast by black voters in a state whose population is
only 15 percent black. The charge is unsupported by the
evidence.
The most sensational "finding" in the majority report, and the
one that received most attention in the press, is the claim that
black voters in the Florida election in 2000 were allegedly
nine times as likely as other residents of the state to have cast
ballots that did not count in the presidential contest, and that
52 percent of all disqualified ballots were cast by black
voters in a state whose population is only 15 percent black.
This charge was naturally newsworthy; however, it is
demonstrably false.
Dr. Lichtman's statistical analysis is badly flawed, strongly
slanted to support preconceived conclusions that cannot
withstand careful scrutiny. The bold assertion that 52 percent
of disqualified ballots were cast by blacks, and that blacks
were nine times as likely to have their ballots rejected as
non-blacks, we will show in detail below, is best described as
nothing more than a wild guesstimate. Dr. Lichtman's other
estimates are not much more reliable, and he fails to examine
the impact of some variables that were of great importance in
determining the outcome.
Below we provide a broader and more sophisticated
regression analysis prepared for us by an econometrician, an
analysis which clashes with that provided in the majority
report on virtually every important point.
Playing Semantic Games: Disfranchisement, Voter Choice, or
Voter Error?
To begin with, disfranchisement is conflated with voter
error. The report talks about voters likely to have their
ballots spoiled; in fact, the problem was undervotes and
overvotes, some of which (particularly undervotes) were
deliberate. But the rest are due to voter error. Or
machine error, which is random, and thus cannot
"disfranchise" any population group. It was certainly not
due to any conspiracy on the part of supervisors of
election; the vast majority of spoiled ballots cast in
counties where the supervisor was a Democrat a point
to which we will return.
The majority report argues that race was the dominant
factor explaining whose votes counted and whose were
rejected. But the method used rests on the assumption
that if the proportion of spoiled ballots in a county or
precinct goes up at the same time that the proportion of
black voters rises, it must be African American ballots
that were disqualified. That conclusion does not
necessarily follow, as statisticians have long understood.
We have no data on the race of the individual voters.
And it is impossible to develop accurate estimates about
how groups of individuals vote (or misvote) on the basis
of county-level or precinct-level averages.
It is important to note at the outset that the majority report's
account of Dr. Lichtman's findings employs language that
serves to obscure the true nature of the phenomenon under
investigation. These pages are filled with references to the
"disenfranchisement" of black voters, as if African Americans
in Florida last year were faced with obstacles comparable to
poll taxes, literacy tests, and other devices by which southern
whites in the years before the Voting Rights Act of 1965
managed to suppress the black vote and keep political office
safely in the hands of candidates committed to the
preservation of white supremacy.
Black votes, we are told again and again, were "rejected" in
vastly disproportionate numbers. "Countless Floridians," the
report concludes, were "denied
their right to vote," and this
"disenfranchisement fell most harshly on the shoulders of
African Americans." In a particularly masterful bit of
obfuscation, the majority report declares that, "persons living
in a county with a substantial African American or people of
color population are more likely to have their ballots spoiled
or discounted than persons living in the rest of Florida." This
alleged fact, the reader is told, "starts to prove the Florida
election was not 'equally open to participation' by all."
Let us be clear: According to Dr. Lichtman's data, some
180,000 Florida voters in the 2000 election, 2.9 percent of
the total, turned in ballots that did not indicate a valid choice
for a presidential candidate and thus could not be counted in
that race. Six out of ten of these rejected ballots (59 percent)
were "overvotes" ballots that were disqualified because
they indicated more than one choice for president. Another
35 percent were "undervotes," ballots lacking any clear
indication of which presidential candidate the voter preferred.
(The other 6 percent were invalid for some other unspecified
reason. Since they are ignored in the majority report, they will
be here as well.)
Hence the chief problem in Florida was voters who cast a
ballot for more than one candidate for the same office, and
the second most common problem was voters who registered
no choice at all. Ballots were "rejected," in short, because it
was impossible to determine which candidate if any
voters meant to choose for president.
Some of these overvotes and undervotes, it should be noted,
may have been the result of deliberate choices on the part of
voters. In fact, Chair Mary Frances Berry remarked at the
hearing in Miami that she herself has sometimes "over-voted
deliberately."
Chair Berry cannot be the only voter in the United States to
make such a choice. According to the exhaustive investigation
of the ballots conducted by the Miami Herald, 10 percent of
all the overvotes in the state showed votes for both Bush and
Gore. Presumably, these were voters attempting to convey
the message that either candidate would be equally
acceptable. Some voters in Citrus County put giants X's
through the names of all presidential candidates, perhaps to
indicate "none of the above."
Similarly, some of the undervotes under discussion here must
been recorded by people who could not settle on a choice
for president but who turned up to register their preferences
in other contests. We know from the Miami Herald's
inspection of the 61,111 undervoted ballots in the state that
almost half 46.2 percent had no markings at all for
president. It seems reasonable to assume that most of them
did not intend to register a choice among the presidential
candidates, and had come to the polls to vote for other
offices.
If these unmarked ballots were produced by voters who
really did not want to make a choice for president, that would
reduce the number of so-called "spoiled ballots" in the state
from 180,000 to 152,000. It would be interesting if we could
make a similar statistical estimate of the proportion of
overvoters who did it deliberately; unfortunately that is
impossible.
What is clear is this: In these instances, overvoting and
undervoting are not "problems" that require "remedies." And
they certainly are not evidence that anyone is being
"disenfranchised." They represent the actual preferences of
the voters in question, and it is misleading to label them
"spoiled" ballots at all.
The majority would have us believe that "countless" numbers
of Floridians who were legally entitled to vote had their
ballots "spoiled." In fact, we are not talking about "countless"
ballots. We are talking about 180,000 invalid ballots, minus
those that did not indicate a clear presidential choice because
the voter had not decided on a presidential preference. Thus
the 180,000 figure, 2.9 percent of the total, is an upper
bound estimate of the true figure, which is undoubtedly
smaller by an unknown amount. The county-by-county
figures on so-called spoiled ballots are likewise
exaggerations, biased upward to an unknown amount.
Still, there are overvotes and undervotes that probably did
not reflect the will of the voters. What accounts for them?
The opening paragraph of the introduction to the majority
report suggests that the issue is whether "votes that were cast
were properly tabulated." What does this mean? Are we to
believe African Americans cast their ballots correctly on
election day, but that the ballots were incorrectly tabulated by
the machines, or the people who conducted manual recounts
in some counties? There is no evidence whatsoever to
support that implication.
Some of the 180,000 rejected ballots may have the result of
machine error, of course but very few. Machine error,
according to experts who have studied it, is rare, involving at
most 1 in 250,000 votes cast. And machine error is obviously
random, and thus cannot "disenfranchise" any population
group. No one has yet shown that a VotoMatic machine can
be programmed to distinguish black voters from others and to
record votes by African Americans in such a way as to
facilitate their rejection.
There is only one other explanation of what the Commission
tendentiously describes as "disenfranchisement." The problem
is voter error, a term that astonishingly appears nowhere in
the majority report. This is the central fact the majority report
attempts to obscure. Some voters simply did not fill out their
ballots according to the instructions. They failed to abide by
the very elementary rule that you must vote for one and only
one candidate for the office of president of the United States,
and therefore their attempt to register their choice failed.
Their ballots were rejected, and their votes did not count.
The majority report argues that race was the dominant
factor explaining whose votes counted and whose were
rejected. But the method used rests on the assumption
that if the proportion of spoiled ballots in a county or
precinct goes up at the same time that the proportion of
black voters rises, it must be African American ballots
that were disqualified. That conclusion does not
necessarily follow, as statisticians have long understood.
We have no data on the race of the individual voters.
And it is impossible to develop accurate estimates about
how groups of individuals vote (or misvote) on the basis
of county-level or precinct-level averages.
The Ecological Fallacy
Did African American voters in the 2000 Florida election
have more difficulty completing their ballots correctly than did
other citizens of the state, and hence have a higher rate of
ballot rejection? Quite possibly so, but Dr. Lichtman's
estimates upon which the Commission relied are open to very
serious doubt. At best, they are highly exaggerated and,
strong evidence (Dr. Lott's research, discussed below)
suggests they are entirely wrong.
How can we figure out whether there were major racial
differences in the rate of voter error or ballot spoilage in the
2000 election? We have no data whatever on the race of
those individuals who cast invalid ballots. We have secret
ballots in the United States, and accordingly cannot know
how any individuals actually voted. Thus we cannot know
with any precision how particular ethnic or racial groups
voted, or at what rate their ballots were actually counted.
Whatever conclusions we draw about the matter must be
based on estimates that will be susceptible to error. The
question is whether the analysis and interpretations offered in
the majority report are at least pretty good approximations of
reality. There are many reasons to doubt that they are.
The majority report attempts to draw conclusions about this
important matter by examining county-level, and to a limited
extent, precinct-level data. It argues that race was the
dominant factor explaining whose votes counted and whose
votes were rejected. The method employed to reach that
conclusion rests on the assumption that if the proportion of
spoiled ballots tends to increase across counties or across
precincts as the proportion of blacks residents in those
counties increases, it must be African American voters whose
ballots were disqualified. This simple methodology may seem
intuitively appealing but it is well established that it is often
wrong.
Statisticians have long understood the difficulty of making
such inferences due to a phenomenon that is known in the
social science literature as the "ecological fallacy." The classic
discussion of this issue is in an article that was published half a
century ago in the American Sociological Review. In that
paper, W.G. Robinson reported that had examined the
correlation between the proportion of a state's population that
was foreign-born and the state's literacy rate. He found,
surprisingly, a positive correlation between the literacy rate
and the proportion of immigrants in the population. Contrary
to the conventional wisdom, the larger the foreign-born
population, the higher the overall literacy rate was in a state.
The correlation was .53, a bit higher than the one found by
Dr. Lichtman between race and ballot spoilage rates.
Did that really prove that Americans born abroad were more
literate, on the average, than those born within the United
States? Robinson chose this case because he had reliable
data against which to check the ecological estimate; census
data were available for individuals. When Robinson
analyzed it, he found that country of birth was negatively
correlated with literacy; the actual figure was -.11.
Immigrants were actually significantly less likely than natives
to be literate, despite the strong state-level correlation
suggesting just the opposite.
The state-by-state correlation gave a completely false picture,
because it happened that the states with highly literate
populations were also more developed economically and
attracted more immigrants because jobs were available there.
New York, for example, was more literate than Arkansas. It
also had a higher fraction of immigrants in its population, but
not enough to pull the state average down very far.
A more recent example derives from the work of an eminent
mathematical statistician at the University of California at
Berkeley, David A. Freedman.
Using data from the 1995 Current Population Survey,
Freedman found that the correlation between the proportion
of immigrants in the population of the 50 states and the
proportion of families with incomes over $50,000 in 1994
was .52. Foreign-born Americans, judging from this
ecological correlation, were considerably more affluent than
their native-born neighbors. But the evidence also allowed
Freedman to look at incomes on the individual level. When
you do that, it turns out that in the nation as a whole, 35
percent of native-born American families were in the $50,000
and over income bracket but only 28 percent of immigrant
families were. The true correlation between being
foreign-born and having a high family income was not the .52
estimated from state-level data; it was instead a mildly
negative correlation of -0.05.
In this instance, too, estimates based on ecological
correlations were not just a bit off, a little imprecise but still
close enough to the truth for most purposes. They were way
off the mark, and indeed had falsely transformed relationships
that were actually negative into positive ones.
The problem of the ecological fallacy afflicts all of the
statistical analyses Dr. Lichtman did for the majority report.
We must remember that counties do not vote. Precincts do
not vote. Only individuals vote. It is impossible to develop
accurate estimates about how groups of individuals vote (or
misvote) on the basis of county-level or precinct-level
averages.
In his appearance before the June 8, 2001 meeting of the
Commission on Civil Rights, Dr. Lichtman sounded a note of
caution about his findings. He declared that a correlation does
not "by itself prove" that there were "disparate rates" at which
ballots by African Americans and "non-African Americans"
were rejected. That is certainly true. But he went on to claim
that the "more advanced statistical procedures" he employed
could reliably do so. Unfortunately, that is not true. The use
of ecological regression techniques does not solve the
problem of the ecological fallacy, because it depends upon
exactly the same aggregated data as simple correlational
analysis, and makes the same, often incorrect, "constancy
assumption." It assumes that there is no relationship between
the composition of geographical areas and the relationship in
question, when in fact there often is.
If the information utilized in an analysis is based on averages
for geographical units, whether they are counties or precincts,
the results will necessarily be imprecise and they may be just
plain wrong, as in the example of immigrant literacy levels
given above. When David Freedman did an ecological
regression of state-level data to assess the relationship
between immigration and family income, he found that it
estimated that fully 85 percent of foreign-born American
families had 1994 family incomes above $50,000. But the
true figure, from individual-level data, was really only 28
percent. Ecological regression, in this case, yielded results
that were wildly mistaken. In another paper, Freedman
provided a similar critique of ecological regression estimates
of political behavior specifically, in instances in which
individual-level data happened to be available, and he found
ecological regression estimates to have been highly unreliable.
In sum, inferences about individual behavior on the basis of
the average distribution of some characteristic across
geographical units are sometimes wildly inaccurate. They
must be examined with great caution and skepticism. The
majority report does not display the necessary caution about
what the facts reveal. A more searching analysis, summarized
below and spelled out in Appendix I, demonstrates how
misleading Dr. Lichtman's findings are.
Factors Other Than Race May Have Explained the
Percentage of Spoiled Ballots The majority's report
assumes race had to be the decisive factor determining
which voters spoiled their ballots. Indeed, its analysis
suggests that the electoral system somehow worked to
cancel the votes of even highly educated, politically
experienced African Americans.
In fact, the size of the black population (by Dr.
Lichtman's own numbers) accounts for only one-quarter
of the difference between counties in the rate of spoiled
ballots (the correlation is .5). But Dr. Lichtman knows
that we cannot make meaningful statements about the
relationship between one social factor and another
without controlling for or holding constant other
variables that may affect the relationship we are
assessing.
The more complex regression analysis that Dr. Lichtman
conducted does not, however, isolate the effect of race
per se from that of other variables that are correlated
with race: poverty, income, literacy, and the like. Or at
least, he fails to provide the details the regression
models essential to understanding his dismissal of
these other factors. And, most important, he never
reports how much of the variance between counties in
the proportion of ballots spoiled can be explained by a
more complex model, such as the one that our own
expert, Dr. John Lott, developed. Our model enables us
to explain 70 percent of the variance (three times as
much as Dr. Lichtman was able to account for) without
considering racial composition at all.
In fact, using the variables provided in the report, Dr.
Lott was unable to find a consistent, statistically
significant relationship between the share of voters who
were African American and the ballot spoilage rate.
Further, removing race from the equation, but leaving in
all the other variables, only reduced the amount of ballot
spoilage rate explained by his regression by a trivial
amount. In other words, the best indicator of whether or
not a particular county had a high or low rate of ballot
spoilage is not its racial composition. Non-racial
information is much more useful.
Was race itself a decisive factor in determining which voters
spoiled their ballots in the 2000 election in Florida, as the
majority report contends? Did the electoral system somehow
work in such a way that even highly educated, politically
experienced African Americans, for example, cast ballots that
were somehow spoiled in some unspecified and mysterious
way? The majority report claims that the answer was yes,
though it provides no indication of how the process worked
to produce that result. Dr. Lichtman's statistical analysis, the
report claims, demonstrates that such was the case.
It does nothing of the sort, even if we set aside for the sake of
argument the serious doubts most statisticians have about the
accuracy of any estimate based on an ecological regression
or correlation. The report begins with the simple correlation
between the percentage of African American registered
voters in Florida's counties and the percentage of spoiled
ballots. That correlation is .50. Speaking in statistical
shorthand, that "explains" 25 percent of the total variance
across the counties. (It doesn't necessarily "explain" anything
in ordinary language, we shall see later). In other words, if
you want to know why some Florida counties have a high
and some a low rate of spoiled ballots, knowing their racial
composition only accounts for one quarter of the difference.
Social scientists know that a simple correlation of about .5
between two variables has very little meaning. We cannot
make meaningful statements about the relationship between
one social factor and another without controlling for or
holding constant other variables that may affect the
relationship we are assessing.
Dr. Lichtman did perform a more complex regression
analysis, so as to isolate the effect of race per se from that of
other variables that happen to be correlated with race, such
as poverty, median income levels, literacy rates and the like.
But neither the account of his work provided in the majority
report nor the more detailed discussion in June 4 technical
report to the commission conform to normal social science
practice. The only regression estimates offered are for the
same two variables used in his simple correlation that
between race and ballot rejection rates with the only
refinement being that he separates undervotes from overvotes
and takes into account the voting system used in each county.
Dr. Lichtman raises the key question: "Is there some other
factor which better explains this disparity in ballot rejection
rates?" But he simply asserts that the answer is no and moves
on, without providing the detail necessary for anyone to know
the statistical basis for his opinion. He offers only two
sentences claiming that applying controls for education do not
weaken the association between race and ballot rejection.
But he never discusses the possible influence of any other
variables, and he never provides the actual regression
models, as is common in reports of quantitative social science
research.
Most striking, and most damaging to the Commission's case,
Dr. Lichtman never reports how much more of the variance
can be explained by using a more complex model that
incorporates other variables.
As we will show below, it is possible to develop a regression
model that explains approximately 70 percent of the variance
in ballot spoilage rates, nearly three times as much as Dr.
Lichtman has been able to account for, without taking the
racial composition of the counties into account at all.
This conclusion derives from an analysis performed at our
request by a first-rate economist, Dr. John R. Lott, Jr. of the
Yale Law School, who was willing to take the time to
evaluate the work of the commission and of Dr. Lichtman,
and even to gather additional data of his own to further
extend the analysis. Lott's report, with accompanying figures
and tables, appears as an appendix to this statement.
Dr. Lott ran a series of regressions, varying the specifications
in an effort to replicate Dr. Lichtman's results. Using all the
variables reported in Appendix I in the majority report, he
was unable to find a consistent, statistically significant
relationship between the share of voters who were African
American and the ballot spoilage rate. The coefficient on the
percent of voters who were black was indeed positive, but it
was statistically insignificant. The chance that the relationship
was real was only 50.3 percent, just about the chance of
getting tails to come up on any one coin toss and far below
the significance level commonly demanded in social science.
Furthermore, when Dr. Lott analyzed the data using a
specification that implied that the share of African American
voters in a county was significantly related to the level of
ballot spoilage, he found that it explained hardly any of the
overall variance. Removing race from the equation but leaving
in all the other explanatory variables only reduced the amount
of ballot spoilage explained by his regression from 73.4
percent to 69.1 percent, only a mere 4.3 percentage point
reduction (see Lott's Table 2 in the attachment).
Indeed, in none of the other specifications provided in his
Table 2 did taking racial information out of the analysis but
leaving in other variables reduce by more than 3 percent the
amount of variance in the spoiled ballot rate that is explained.
Consequently, it simply is not true that the best indicator
of whether or not a particular county had a high or low
rate of ballot spoilage is its racial composition. One can
predict that with a much higher degree of confidence by
looking at other, non-racial information.
Was Education the Problem?
The obvious explanation for a high number of spoiled
ballots among black voters is their lower literacy rate.
Dr. Lichtman has a cavalier discussion of the question,
and his conclusion that literacy rates were irrelevant
makes no sense. (In fact, the report itself recommends
"effective programs of education for voters
")
Moreover, the data upon which he relies are too crude to
allow meaningful conclusions they are not even
broken out by race.
Although it does not take a high level of literacy to follow the
instruction, "Vote for ONE of the following," or "Fill in the
box next to the name of the candidate you wish to vote for," it
does take some reading ability. We know that some
Americans today, regrettably, find it extremely difficult to
understand even the simplest written instructions. And,
unfortunately, this group is disproportionately black. The U.S.
Department of Education's 1992 Adult Literacy Study found
that 38 percent of African Americans but only 14 percent
of whites ranked in the lowest category of "prose literacy,"
which was defined as being unable to "make low-level
inferences based on what they read and to compare or
contrast information that can easily be found in [a] text."
Black Americans, the study found, were 2.7 times as likely as
whites to have the lowest level of literacy skills. Likewise, the
1998 National Assessment of Educational Progress found
that 43 percent of African American 12th-graders had
reading skills that were "Below Basic," as compared to just
17 percent of whites. Black students were 2.5 times as likely
as whites to lack elementary reading skills. Among adults
employed full-time, blacks are 4.1 times more likely than
whites to be in the lowest prose literacy category.
National studies provide no data on Florida specifically.
However, we know from the National Assessment of
Educational Progress that black 4th- and 8th-graders in
Florida (no state-level data is available for 12th-graders) are
no better readers than their counterparts elsewhere. Indeed,
their scores are below the national average for African
Americans. No fewer than 57 percent of Florida's black
8th-graders in 1998 were Below Basic in reading, 10 points
above the national average for African Americans, and 2.7
times as high as the white figure.
The majority report, though, denies that racial differences in
literacy levels could be the source of the problem. It devotes
only a brief paragraph to the matter, claiming that "a multiple
regression analysis that controlled for the percentage of high
school graduates and the percentage of adults in the lowest
literacy category failed to diminish the relationship between
race and ballot rejection."
But the regression results themselves are not provided for the
critical reader to assess. When one turns to Dr. Lichtman's
actual report for greater illumination, one finds nothing more
than the exact language used in the commission report. This is
a cavalier way to treat an issue as serious as this one.
The claim that the incidence of ballot spoilage or voter error
is unrelated to education is counter-intuitive. It is also
extremely puzzling, because just a few pages later in the same
chapter the report addresses possible solutions to the
problem. It urges the adoption of optical scanning systems
with immediate feedback to voters throughout Florida, but
then goes on to say that this would not "eliminate the disparity
between the rates at which ballots cast by African Americans
and whites are rejected." It estimates that it would only cut
the disparity by about half. What else could be done? The
Commission's answer is "effective programs of education for
voters, for election officials, and for poll workers."
The commission majority seems to be declaring both that:
1.The lower average level of literacy among Florida's
blacks has nothing to do with the allegedly higher rate
of voter error by blacks; and
2.The solution to this problem is for the state of Florida
to launch a huge new program designed to educate
black voters on how to vote successfully.
The logic eludes us.
Dr. Lichtman's attempt to assess the role of education is
cursory, and the data upon which he relies is too crude to
allow meaningful conclusions. The "synthetic estimates of
adult literacy proficiency" he uses have wide confidence
intervals an average of 6 percent. More important, the
literacy data Dr. Lichtman used in his analysis are not broken
down by race. So they cannot tell us anything about whether
the small fraction of a county's voters who failed to cast a
ballot successfully were people who had difficulty reading and
what the racial composition of that group might be.
Remember that the highest rate of ballot spoilage in any
county was 12.4 percent, and that it was below 5 percent in
nearly two-thirds of the counties. So we are talking about a
very small group, and one whose presence is not likely to
show in county-wide averages.
Palm Beach County, for example, led the state in the number
of spoiled ballots nearly 30,0000. Some 6.4 percent of all
the ballots cast there were invalid. The proportion of Palm
Beach residents who ranked in the bottom literacy category
was 22 percent, a little below the state average of 25
percent. And the proportion who had attended college was
48 percent, again above the state average. But this does not
allow us to conclude that the 6.4 percent of Palm Beach
voters who failed to complete their ballots successfully were
not primarily people who had difficulty in reading,
comprehending and following ballot instructions.
| How Many of the Spoiled Ballots Were Cast by First-time Voters? |
An important source of the high rate of ballot spoilage in
|