In Dissent
                 A response to the Florida election report.

        By Commissioners Abigail Thernstrom & Russell G. Redenbaugh
 

                   The United States
                  Commission on Civil Rights,
                  charged with the statutory duty
                  to investigate voting rights violations in a fair and objective
                  manner, has produced a report that fails to serve the public
                  interest. Voting Irregularities Occurring in Florida During
                  the 2000 Presidential Election is prejudicial, divisive, and
                  injurious to the cause of true democracy and justice in our
                  society. It discredits the Commission itself and substantially
                  diminishes its credibility as the nation's protector of our civil
                  rights. 

                  The Commission's majority report is a partisan document that
                  has little basis in fact. Its conclusions are based on a deeply
                  flawed statistical analysis coupled with anecdotal evidence of
                  limited value, unverified by a proper factual investigation. This
                  shaky foundation is used to justify charges of the most serious
                  nature — questioning the legitimacy of the American electoral
                  process and the validity of the most recent presidential
                  election. The report's central finding — that there was
                  "widespread disenfranchisement and denial of voting rights" in
                  Florida's 2000 presidential election — does not withstand
                  even a cursory legal or scholarly scrutiny. Leveling such a
                  serious charge without clear justification is an unwarranted
                  assault upon the public's confidence in American democracy. 

                  Using all the variables in the statistical analysis in the majority
                  report, Dr. John Lott, an economist at Yale Law School, was
                  unable to find a consistent, statistical significant relationship
                  between the share of voters who were African Americans
                  and the ballot spoilage rate. 

                  Furthermore, Dr. Lott conducted additional analysis beyond
                  Lichtman's parameters, looking at previous elections,
                  demographic changes, and rates of ballot spoilage. His
                  analysis found little relationship at all between racial
                  population change and ballot spoilage, and the one
                  correlation that is found runs counter to the majority report's
                  argument: An increase in the black share of the voting
                  population is linked to a slight decrease in spoilage rates,
                  although the difference is not statistically significant. 

                  Nothing is more fundamental to American democracy than
                  the right to vote and to have valid votes properly counted.
                  Allegations of disfranchisement are the fertile ground in which
                  a dangerous distrust of American political institutions thrives.
                  By basing its conclusion on allegations that are driven by
                  partisan interests and which lack factual basis, the majority on
                  the Commission has needlessly fostered public distrust,
                  alienation and manifest cynicism. The intent of the partisan
                  conclusions of the report is to label the outcome of the 2000
                  election as illegitimate, thereby calling into question the most
                  fundamental basis of American democracy. 

                  Obvious partisan passions not only destroyed the credibility
                  of the report itself, but informed the entire process that led up
                  to the final draft. At the Florida hearings, Governor Jeb Bush
                  was the only witness who was not allowed to make an
                  opening statement. The Chair, Mary Frances Berry, was
                  quoted in the Florida press as comparing the Governor and
                  Secretary of State to "Pontius Pilate…just washing their
                  hands of the whole thing." On March 9, six commissioners
                  voted to issue a "preliminary assessment" — in effect, a
                  verdict — long before the staff had completed its review of
                  the evidence. 

                  The statistical analysis upon which many of the final report's
                  findings are based was conducted by an historian with close
                  ties to Albert Gore, Jr. The report claims that "affected
                  agencies were afforded an opportunity to review applicable
                  portions"; in fact, affected parties were never given a look at
                  the preliminary assessment, and had only ten days to review
                  and respond to the final report, in violation of established
                  procedures and previous promises. Our memoranda to the
                  chief of staff throughout the process regularly went
                  unanswered. 

                  Most recently, a request for basic data to which we — and
                  indeed, any member of the public — were entitled was
                  denied to us. The Commission hired Professor Allan
                  Lichtman, an historian at American University, to examine the
                  relationship between spoiled ballots and the race of voters.
                  We asked for a copy of the machine-readable data that
                  Professor Lichtman used to run his correlations and
                  regressions. That is, we wanted his computer runs, the data
                  that went into them, and the software he used. Obviously, he
                  could have easily given that to us. The Commission had the
                  temerity to tell us that it did not exist — that the data as he
                  organized it for purposes of analysis was literally unavailable.
                  Professor Lichtman, who knows that as a matter of scholarly
                  convention such data should be shared, also declined to
                  provide it. Evidently, Dr. Lichtman and the majority on the
                  Commission have no confidence in their own numbers and
                  analysis. 

                  Process matters. And that is why it is important to examine,
                  with integrity, violations of the electoral process in Florida
                  and other states. When the process is right, participants on
                  another day can revisit the outcome — use the procedures
                  (fair and thus trusted) to debate policy or to vote again. But
                  when the process is corrupt, the conclusions themselves
                  (current and future) are deeply suspect. The Commission
                  investigated procedural irregularities in Florida; it should have
                  gotten its own house in order first. 

                  Had the process been right, the substance might have been
                  much better. The Commission's staff would have received
                  feedback from Florida officials, commissioners, and other
                  concerned parties, on the basis of which it might have revised
                  the report. It should be consulting with commissioners in the
                  course of drafting a report, including those who do not share
                  the majority view. As it is, at great expense, the Commission
                  has written a dangerous and divisive document. And thus it
                  certainly provides no basis upon which to reform the electoral
                  process in Florida or anywhere else. 

                  In the pages below, we will argue: 

                  I. 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 the Florida election in 2000 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. 

                  (a) Disfranchisement is not the same as voter error. The
                  report talks about voters likely to have their ballots spoiled; in
                  fact, the problem was undervotes and overvotes, some of
                  which were deliberate (the undervotes, particularly). 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 elections; the vast majority of spoiled ballots
                  were cast in counties where the supervisor was a Democrat. 

                  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 is higher
                  in places with a larger black population, 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. 

                  (b) 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). However it is clear 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 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
                  developed by our own expert, Dr. John Lott of the Yale Law
                  School. 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 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. 

                  (c) The obvious explanation for a high number of spoiled
                  ballots among black voters is their lower literacy rate.
                  Dr. Lichtman offers only 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 broken down by race, for one
                  thing. 

                  (d) First time voters: An important source of the high rate of
                  ballot spoilage in some Florida communities may have been
                  that a sizable fraction of those who turned out at the polls
                  were there for the first time and were unfamiliar with the
                  electoral process. Impressionistic evidence suggests that
                  disproportionate numbers of black voters fell into this
                  category. The majority report's failure to explore — or even
                  mention — this factor is a serious flaw. 

                  (e) The Time Dimension: Most social scientists understand
                  that the interpretation of social patterns on the basis of
                  observations at just one point in time is dangerously simplistic.
                  But that is all the majority report offers. It focuses entirely on
                  the 2000 election returns. Dr. Lott did two analyses that take
                  the time dimension into account. 

                  He looked at spoilage rates by county for the 1996 and 2000
                  presidential races, and compared them with demographic
                  change. A rise in a county's black population did not result in
                  a similar rise in spoilage rates, suggesting, again, that race is
                  not the explanatory factor. 

                  He also examined data from the 1992, 1996, and 2000
                  races, and found that the "percent of voters in different race
                  or ethnic categories is never statistically related to ballot
                  spoilage." 

                  (f) County-level data v. Precinct data: The majority report,
                  as earlier noted, speaks of black ballots as nine time more
                  likely to be spoiled than white ballots. And it presents some
                  precinct-level data, providing estimates based on smaller units
                  that are likely to be somewhat closer to the truth than
                  estimates based on inter-county variations. Dr. Lichtman's
                  own numbers show that county-level and precinct-level data
                  yielded quite different results. Ballot rejection rates dropped
                  significantly when the precinct numbers were examined, even
                  though looking at heavily black precincts should have
                  sharpened the difference between white and black voters,
                  rather than diminishing it. Dr. Lichtman obscures this point by
                  shifting from ratios to percentage point differences. 

                  Dr. Lichtman's precinct analysis is just as vulnerable to
                  criticism as his county-level analysis. It employs the same
                  methods, and again ignores relevant variables that provide a
                  better explanation of the variation in ballot spoilage rates. 

                  (g) Who Is Responsible for Elections? The majority report
                  charges "disenfranchisement" and lays the blame at the feet of
                  state officials — particularly Governor Jeb Bush and
                  Secretary of State Kathryn Harris. In fact, however, elections
                  in Florida are the responsibility of 67 county supervisors of
                  election. And, interestingly, in all but one of the 25 counties
                  with the highest spoilage rates, the election was supervised by
                  a Democrat — the one exception being an official with no
                  party affiliation. 

                  Dr. Lott added another variable to the mix: the race of the
                  election supervisor. And he found that having Democratic
                  officials in charge increases the ballot spoilage rate
                  substantially, but the effect is even stronger when that
                  Democratic official is African American. Obviously no
                  officials were out to disfranchise black voters, and the
                  correlation points once again to the limitations of ecological
                  regressions. 

                  The majority report argues that much of the spoiled ballot
                  problem was due to voting technology. But Democratic Party
                  officials decided on the type of machinery used, including the
                  optical scanning system in Gadsden County, the state's only
                  majority-black county and the one with the highest spoilage
                  rate. 

                  (h) Hispanics have been mostly forgotten. Hispanics are a
                  protected group under the Voting Rights Act. Moreover, the
                  majority report speaks repeatedly of the alleged
                  disenfranchisement of "minorities" or "people of color." One
                  section is headed "Votes in Communities of Color Less
                  Likely to be Counted." And yet the crucial statistical analysis
                  provided in Chapter 1 entirely ignores Florida's largest
                  minority group — people of Hispanic origin. The analysis in
                  the Commission's report thus excluded more Floridians of
                  minority background than it included. 

                  The analysis conducted by Dr. Lichtman treats not only
                  Hispanics but Asians and Native Americans as well as if they
                  were, in effect, white. He dichotomizes the Florida population
                  into two groups, blacks and "nonblacks." 

                  In the revised report, Dr. Lichtman did add one graph dealing
                  with Hispanics in the appendix, but this addition to his
                  statistical analysis is clearly only an afterthought. At the June
                  8th Commission hearing Dr. Lichtman stated he looked at this
                  issue only at the last minute (literally the night before).
                  Obviously, his primary analysis ignored Hispanics. 

                                          (continued on page two)

                  1 Report, 154
                  2 Report, 18.
                  3 Report, 21. Note that later in the report, on page 148, the majority
                  asserts that it was highly anomalous that 63 percent of spoiled
                  ballots in Palm Beach County were overvotes, and blames the alleged
                  anomaly on the infamous butterfly ballot. The pattern, according to
                  the report, was "just the opposite of what we normally observe,
                  which is five percent or less of the spoiled ballots." How could the
                  author of this passage possibly think that 5 percent or less was the
                  norm for overvotes in Florida when the Lichtman cited earlier reveal
                  earlier show that fully 59 percent of all the spoiled ballots in the state
                  were overvotes 4 Martin Merzer, The Miami Herald Report:
                  Democracy Held Hostage (New York: St. Martin's Press, 2001), 194
                  5 Ibid., 195.
                  6 Ibid., 230-231
                  7 Report, 1
                  8 According to the Caltech/MIT Voting Project, "state and federal
                  voting machine certifications tolerate very low machine failure rates:
                  no more than 1 in 250,000 ballots for federal certification and no more
                  than 1 in 1,000,000 in some states." The problem, according to these
                  investigators, has to do with "how people relate to the
                  technologies...." See the Caltech/MIT Voting Project, "A Preliminary
                  Assessment of the Reliability of Existing Voting Equipment,"
                  February 1, 2001, 13.
                  9 Exit polls are commonly used to estimate how particular groups
                  voted, and even they are far from perfect. One flaw is that absentee
                  voters are not represented at all. In any event, we can't tell from an
                  exit poll whether someone failed to complete a valid ballot; if they
                  thought they had erred, presumably they would have had it
                  invalidated and received another.
                  10 W.G. Robinson, "Ecological Correlations and the Behavior of
                  Individuals," American Sociological Review, vol. 15 (June, 1950),
                  351-357.
                  11 D.A. Freedman, "Ecological Inference and the Ecological Fallacy,"
                  University of California at Berkeley Department of Statistics
                  Technical Report No. 549, Oct. 15, 1999, This paper will appear as a
                  chapter in the forthcoming International Encyclopedia of the Social
                  Sciences. 
                  12 Transcript of June 8, 2001 meeting, 42. 
                  13The explanation is that immigrants tend to be attracted to the richer
                  states--California and New York rather than Tennessee and
                  Mississippi. Thus their presence is associated with high average
                  incomes at the state level, but that does not mean that their average
                  incomes are especially high.
                  14 D. A. Freedman, S. P. Klein, M. Ostland, and M. Robert, "On
                  'Solutions' to the Ecological Inference Problem," Journal of the
                  American Statistical Association, vol. 93 (December 1998),
                  1518-1523.
                  15 Report, 21,
                  16 Lichtman, "Draft Report on the Racial Impact of the Rejection of
                  Ballots Cast in the 2000 Presidential Election in the State of Florida,"
                  June 4, 2001. 
                  17 Lott, "Issues in the Interpretation of the Statistical Evidence
                  Employed in the Report of the U.S. Commission on Civil Rights on
                  the 2000 Election in Florida," 3.
                  18 National Center for Education Statistics, Adult Literacy in
                  America: A First Look at the Results of the National Adult Literacy
                  Survey, National Center for Education Statistics (Washington, D.C.:
                  U.S. Government Printing Office, 1993), 18, 113. 
                  19National Center for Education Statistics, NAEP 1998 Reading
                  Report Card for the Nation and the States, NCES 1999-500
                  (Washington, D.C.: U.S. Department of Education, 1999), 70. 
                  20 National Center for Education Statistics, Literacy in the Labor
                  Force: Results from the National Adult Literacy Survey, NCES
                  1999-470 (Washington, D.C.: U.S. Department of Education, 1999), 57.
                  21NAEP 1998 Reading Report Card, 260, and data from the NAEP
                  website.
                  22 Report, 22; Lichtman Report, 6.
                  23 Report, 34.
                  24 CSAS website
                  25 Transcript of June 8, 2001 Meeting, 44.
                  26 Ibid, 44.
                  27 Report, 141
                  28 U.S. Census Bureau, Profiles of General Population
                  Characteristics, 2000 Census of Population and Housing: Florida,
                  May 2001, Table DP-1. We state that the black population was
                  approximately 15 percent of the total because its exact size depends
                  upon the definition you use. Some 14.6 percent of Floridians reported
                  that their sole race was black. If you add in people who considered
                  themselves both black and something else, the figure increases to
                  15.5 percent, still substantially smaller than the Hispanic population.
                  29 Ibid. In addition to the 2.7 million Hispanics and the 450,000
                  Asians or American Indians, another 697,000 Floridians reported that
                  they were of "other race," meaning other than white, black, American
                  Indian, Asian, or Pacific Islander. Most of these "other race"
                  respondents were, in all likelihood, Latinos, and thus cannot be fairly
                  added to the total excluded from attention because it would entail
                  double counting. All Hispanics were excluded, however they
                  answered the race question. 
                  30 Transcript of United States Commission on Civil Rights meeting,
                  Washington, D.C., June 8, 2001, 46.
                  31 http://www.american.edu/cas/faculty.shtml#HISTORY. WMA
                  32Transcript of U.S. Commission on Civil Rights hearing,
                  Tallahassee, Florida, January 11, 2001, PAGE TK