Mebane & Dill Uncover Very Suspicious Undervote Pattern in Sarasota Cnty, FL 2006 Election

From: Kathy Dopp <kathy_dot_dopp_at_gmail_dot_com>
Date: Wed Jan 24 2007 - 19:46:24 CST

Perhaps you OVC technical wizards will have some possible explanations
for this voting pattern?

New study of Sarasota County by Mebane and Dill find a group of ballots where:

1. undervote pattern is related to a particular error message seen on
voting machines, and

2. undervote pattern is related to partisan voting pattern on ballots

Sounds consistent with vote tampering to me, doesn't it? But Dill
and Mebane do not make that conclusion.

Factors Associated with the Excessive CD-13 Undervote in the 2006
General Election in Sarasota County, Florida
Walter R. Mebane, Jr.† David L. Dill‡
Initial draft: 18jan2007; this draft: 23jan2007

Based on statistical analysis of detailed electronic ballot and event
log data from the Sarasota general election ending November 7, 2006,
we find that none of the many theories advanced so far to explain the
extraordinarily high undervote rate in the Florida Congressional
District 13 (CD-13) race adequately account for systematic
covariations between that undervote and other identifiable factors,
such as voting patterns in other contests and unusual events on the
voting machines. At this time, we are unable to propose a convincing
mechanism based on voter, machine or ballot characteristics that
completely explains the phenomenon.

This paper describes the data and statistical analysis we have
performed and evaluates explanations that have been advanced by others
as well as plausible explanations that we propose in light of our
analysis. Our results are suggestive but in important respects
puzzling. In a nutshell, the excessive CD-13 undervote rate in
Sarasota County is not yet well-understood, and will not be understood
without further investigation. On its own, further statistical
analysis of the kind of data we examine here probably cannot explain
the undervotes.

Even though important components of the CD-13 undervote rate are
readily explicable, the available explanations do not fully account
for the phenomenon. Some factors that correlate with the CD-13
undervote connect to simple and general plausible explanations.
Several hundred CD-13 undervotes come from voters who, based on votes
they cast for other offices, seem to be disinclined to vote for
candidates affiliated with either the Democrat or Republican party:
the CD-13 race did not offer a third-party alternative and did not
allow write-in votes. Several thousand CD-13 undervotes appear on
ballots that also have undervotes for other offices, notably for the
five statewide offices that appeared on Florida ballots in 2006. It is
plausible that most of these undervotes reflect voters who simply did
not care to vote for several offices. These undervotes are not related
to any voting machine characteristics we were able to observe.

Other correlates of the CD-13 undervote rate are straightforward to
describe, but the general explanations that may connect to them are
not so clear. We find differences of hundreds of CD-13 undervotes when
we compare voting machines that have different observable
characteristics. Hundreds of these undervotes are related to a
specific error message in the event log file (the event log file
supposedly reports every transaction that occurred on each voting
machine). The CD-13 undervote rate varies substantially across voters,
differences that correlate significantly with the partisan balance
among the votes recorded for the five statewide offices. Ballots that
have even one additional statewide office with a vote for a Democrat
rather than for a Republican tend to have higher rates of CD-13

The principal question we cannot answer is whether these patterns
reflect voluntary behavior or artificial errors or manipulations. The
urgency of this question is highlighted by the fact that the
relationship between the CD-13 undervote rate and the statewide office
voting pattern differs depending on whether a particular error message
occurs on the voting machine on which the votes were cast. In the data
for votes cast on election day, this interaction effect is
statistically significant. The same effect is observable but not
significant in the data for votes cast during early voting. The number
of CD-13 undervotes that appear to be directly implicated in this
interaction effect is relatively small, but it is worrisome to see any
sign that an error in the voting machines' operation is a marker for
otherwise comparable voters having their votes recorded differently.

Moreover it is difficult to conclude that persistent differences in
basic propensities to undervote explain the differences in CD-13
undervoting among ballots that have different statewide voting
patterns, because we find that the same subsets of ballots have
substantially different undervote rates for other offices. For
instance, the CD-13 undervote rate is higher on ballots that have
votes for Democrats for all five statewide offices than it is on
ballots that have all five votes for Republicans, but it is the
straight-Republican ballots that have the higher undervote rate in
votes for Hospital Board Southern District Seat 1. The relationship
between partisan voting for the statewide offices and undervoting is
peculiar to the CD-13 race.

Such peculiarity raises a substantial question about the popular idea
that the large CD-13 undervote was caused by the format of the ballot
(Mahlburg and Tamman 2006; Tamman and Doig 2006; Frisina, Herron,
Honaker, and Lewis 2006). Clearly the ballot format caused many of the
CD-13 undervotes, but it is not at all clear how many. The ballot
format cannot explain the differences we observe between voting
machines and between subsets of ballots. The ballot format cannot
explain why the distribution of undervotes differs significantly
between the race for Hospital Board Southern District Seat 1 and the
CD-13 race. The alternatives for the Hospital Board Southern District
Seat 1 race appear on the ballot with notable features that are very
similar to those that are often credited with causing the problems
with the CD-13 race.

OVC-discuss mailing list
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
Received on Tue Jan 1 14:12:48 2008

This archive was generated by hypermail 2.1.8 : Tue Jan 01 2008 - 14:12:51 CST