California Recall Wrap-Up ~ #31 By Alan F. Kay, PhD © 2003, (fair use with attribution and copy to author) Dec. 10, 2003 By September 2003, the California courts resolved an enormous range of contentious opinions on how the special election to recall and replace Gov. Gray Davis was to be conducted.  The election would proceed with this firm schedule:    -- election day: Oct. 7    -- certification of the vote by Secretary of State Kevin Shelley, no later than: Nov. 15    -- if Davis was recalled, the inauguration date of the new governor: Nov. 17.  The ballot, it was agreed, was to have two parts.  In Part 1, the voter was asked to vote "yes" (Gray Davis should be recalled) or "no" (he should not be).  If, of those voting in Part 1, 50% or more voted “no,” Davis would serve out his remaining three-year term.  If the number of “no” voters in Part 1 was less than 50%, the leading candidate in Part 2 would be governor-elect.  Davis was prohibited from running in Part 2.  Mindful of the imperative set by this tight schedule, draconian measures were put in place:  no judicial challenge of the election outcome and no request for a recount.  The election could easily have turned out to be a fiasco.  Everybody recognized that the votes cast to retain Davis as governor ("no" in Part 1), might well be less than 50% of the votes cast in Part 1, but still much greater than the votes cast for the leading candidate in Part 2.  If public appeal were broadly distributed among the 163 candidates of Part 2, this might easily have resulted in a governor-elect whose vote total was less than 50%, possibly even less than 10%, of the votes cast for Davis in Part 1.  Though far-fetched, an even more disastrous outcome would have been possible.  Davis, just shy of 50% of the vote in Part 1, might have had more votes cast favoring him than all the votes cast for all the candidates in Part 2, simply because a little over half of the voters who voted in Part 1 failed to vote in Part 2. Those who cast a vote in Part 1 in the actual election were greater in number than those who cast a vote in Part 2, by an amount that turned out to be much smaller than 50 percent. Expressed as a percent of the total votes cast in part1, it was a mere 3.8%. Of the votes cast in Part 2, Arnold Schwarzenegger captured a healthy plurality, 48.6%.  It is reasonable to assume that the voters of California realized that in order to have a popularly elected governor rather than someone selected with little public approval as the result of a ridiculously quirky ballot process, they would somehow have to be in good agreement on who should replace Davis.  Somehow, they were.  After election day, political analysts and pundits explained to the world how, all things considered, Schwarzenegger really was quite a good choice, but none explained how the public had come together, only in the last week of the election, in apparent agreement on voting for his candidacy.  Shelley and the 58 California county supervisors of elections were saved from a real fiasco by what could be called the "wisdom of the people".  They should be grateful. The officials had to make sure, county by county, that on short notice there would be enough voting machines to handle what could be, and indeed was, a large turn-out.  Eleven different makes and models of machines were certified as acceptable by Shelley, and each county quickly had to acquire and deploy a huge number of them and train precinct recruiters how to guide voters through the process.  The officials received the support of an organization www.votewatch.org  that trained volunteers on proper voting procedures to assist at precincts throughout the state, in preparation for going national with a similar service for the 2004 elections. The Office of the Secretary of State developed a magnificent central website (www.ss.ca.gov ).  One lengthy page gave the characteristics and specifications of the certified voting machines and data on how they were arrayed county by county.  Each county supervisor of elections, starting at the close of the election on Oct. 7, reported to the central website the current available totals of votes for both parts of the ballot (and two referenda also on the ballot following Part 2) --  including votes cast in three unusual categories in a manner never before used in a large election.  Any voter could vote at any precinct even if the voter was not accepted as being registered at that precinct, and such votes would be tallied in a "provisional" category in the county.  "Unprocessed absentee" ballots were kept in an "absentee" category.  A third category, labeled "destroyed ballots", was fortunately inconsequentially small.  When a provisional voter offered acceptable proof that s/he indeed should have been allowed to vote, that vote was added to the "provisional" category displayed on the appropriate page of the central website.   Absentee and destroyed ballots were handled in a similar manner. The total votes for the yes/no question of Part 1 and for each of the 163 candidates of Part 2, were displayed automatically updated every few minutes, 24 hours a day, from Oct. 7 to Nov. 14.  Any media observer -- or interested person around the world -- could access the website and see these latest totals slowly climb over the 38 day period, while   percentages of the total vote changed amazingly little, less than ±0.1%.  The final votes were certified by Shelley in a way strange to all of us familiar with statewide polls, especially elections that got as much media coverage as this one.  Though virtually everybody believed that Schwarzenegger was the governor-elect, nowhere did the certification state that.  One could deduce who was governor-elect from looking at the data.  The percentages showed that Schwarzenegger's total exceeded that of Davis and the other leading candidates.  But nowhere to be seen was the total number of votes of all 163 candidates, nor accurate percentages for many minor candidates.  It is almost as if Shelley had told those preparing the certification documents, "Don't add up the totals or figure the percentages for all the minor candidates.  They're not necessary and you might make a mistake.  Those numbers could not change the outcome.  No one could lodge a significant complaint.  Don't bother getting them accurate."  There was no bias visible in Shelley's operation or documentation. What a contrast to Florida's secretary of state in 2000, Katherine Harris who twice on TV joyfully announced George W. Bush the winner in Florida by a paper thin margin and thereby nationally by a single electoral vote.  Harris was biased by her widely-known Republican activism in the campaign and made no apologies for these election follies that were her department's responsibility:     --  The butterfly ballot, which did not link the name in a straight line to the check-box for the candidate.      --  Printed advice on ballots that if followed would void the ballot.  Example, "Be sure to vote on every page."      --  Over 40,000 ballots cast but mysteriously never counted.     --  Registered voters prevented from voting when incorrectly told that they were not registered.     -- A massive campaign conducted by Harris to remove thousands of names from voting lists throughout Florida for dubious reasons, e.g. having names similar to the name of someone who might have once been convicted of what was a misdemeanor, but later made a felony under Florida law.    After going to enormous lengths to run an unbiased election with a full 38 days to get the votes properly counted and the counts announced accurate to a single vote, Shelley had made an enormous and successful effort to avoid the Harris follies. The last thing that Shelley or anyone else in California of either party wanted was a repeat production.   The fundamental flaw of the California recall election was the two-part ballot.  Of those voting in Part 1, 44.6% favored Davis.  Of those voting in Part 2, 48.6% favored Schwarzenegger, and Schwarzenegger was the second choice of any voter who had already voted for Davis in Part 1.  If 198,465 or more voters voted "no" in Part 1 and for Schwarzenegger in Part 2, then it would be true that based on first choice only, Davis had more votes than Schwarzenegger.  If this were true, then, in the name of majority rule in a democracy, Davis should have been the winner and remained in office.   The magic number, 198,465, was small, only 5.0% of the total pro-Davis vote.  A full analysis, such as presented in "California Inherits Florida Mantle" on the website of The Polling Critic, Nov. 22, 2003, shows that it is almost certain that the true desire of the majority of voters was to retain Davis as governor.  The tiny uncertainty could easily have been eliminated by any recount that found 198,465 or more of those voting "no" in Part 1 and "Schwarzenegger" in Part 2.  The rules laid down by the California courts prevented the recount, and thereby ignored the will of the majority and produced a failure of democracy.  These conclusions are unaffected by the fact that from Oct. 8 on, probably no one had expected that Davis would remain as governor, even those few who may believe, as I do, that Davis was the voters' first choice.  Three aspects of the ballots: design, instructions and layout for both California 2003 and Florida 2000 were bad and thwarted the will of the people.  Thus, despite all the effort to avoid a Florida-like fiasco, California has been tarred by the same brush.  To be fair to the individuals responsible, Harris should be covered with tar from head to toe; and Shelley, well, should have a little dab on his new white suit. Why do I see things this way and almost nobody else does?  I have had years of experience, often in collaboration with the best pollsters in the United States, designing, conducting and analyzing polls where the public was asked to choose the most favored from a bunch of choices offered.  The choices could be for the favorite among policy proposals, election candidates, or whatever.  The outcome is most reliable if the choices are all offered in the same frame. A simple example illustrates why.  Suppose a poll offers just three response choices, A, B, and C, which happen to be related in a not uncommon way, known as "cyclical preference."  This means that A is preferred over B, B is preferred over C, and C is preferred over A.  If two questions are used, the first, A matched against B; and the second, the preferred A of the first question matched against C, then C becomes the most preferred of the three.  Or put more generally, the choice that is tested only in the second question always wins.  This, of course, is paradoxical, and impractical nonsense.  The simple solution is to ask the public with all choices in a single question, "Which in a three-way match (and a level playing field) is preferred, A, B, or C?"  The public will easily make its choice.  The California recall election, commendable as it was in many ways, should not have had two parts. o - o - o - o - o - o - o - o Alan F. Kay is a mathematician, social scientist and pioneer of public-interest polling. He has authored Locating Consensus for Democracy and numerous public policy articles and holds several patents. (see www.publicinterestpolling.com)

Public-interest polling - what's that?    Top    Home    Search    CDI Home