Milwaukee Journal Sentinel, February 15, 2005

By Greg J. Borowski and Tom Kertscher

An independent firm has found flaws in the Nov. 2 election records of many Wisconsin communities that echo some of the problems identified by the Journal Sentinel in Milwaukee.

Wisconsin Voter Lists, which is updating the files it uses to sell voter lists to political candidates, found several communities with significant gaps between the number of people listed as having voted and the number of votes cast in the presidential contest.

The firm, which has collected data from more than 50 Wisconsin communities so far, shared its information with the Journal Sentinel, which first identified a 7,000-vote gap in Milwaukee. Several communities showed gaps that, on a percentage basis, rivaled the city's disparity.

That includes Madison, Fond du Lac, Neenah and Eau Claire. Typically, though, gaps were much smaller and many communities showed no gaps at all, underscoring the importance of a vote-to-voter match in even high-turnout elections.

In some cases, election officials said the problem was due to an incomplete voter list they sent to the firm. In others, though, they were surprised to learn that their voter records had problems - including discrepancies that, in a few spots, listed more voters than votes.

In Neenah's computer files, the firm found 12,637 people listed as having voted, but 13,226 ballots cast in the presidential race. That means 95.5% of the votes were accounted for.

The city provided the data before discovering that a computer glitch caused more than 500 voter histories not to be updated, City Clerk Patty Peterson said. That alone might explain the gap of nearly 600 reported by the firm because the gap is now fewer than 10, she said.

"We figure that our records are pretty cleaned up," Peterson said.

Last month, the Journal Sentinel identified the 7,000-vote gap in Milwaukee as it reviewed records from the Nov. 2 election. The gap represents the difference between the number of people listed by the city as having voted (269,212) and the total number of ballots cast (277,535).

That difference is more than 8,300, but the gap is closer to 7,000, because the city has 1,300 same-day registration cards it could not process. Officials blamed the gap on problems in entering polling place logbook information into the city computer system.

The newspaper, however, has found similar gaps in the logbooks themselves, which critics say is more troubling.

Record-keeping woes
The review of other Wisconsin cities reflects the first sort of discrepancy - one related to how well voters are recorded in the computer in the weeks after the election.

It cannot be used to determine whether any polling-place discrepancies exist in any of the communities, though calls to many local election officials tend to discount that likelihood.

The information was compiled by the Michigan-based firm Practical Political Consulting, which operates in the state as Wisconsin Voter Lists. In a similar fashion to what the Journal Sentinel did with Milwaukee's vote totals, it compared the number of names listed as voting in each community to the number of votes counted in the Nov. 2 presidential race.

For Milwaukee, it showed 267,344 people voting and 276,921 votes cast, or 96.5% of the votes accounted for in the city's own computer system.

The numbers are slightly different than those used by the newspaper, because the firm deleted obvious duplications among voter listings and used the total of votes in the presidential race. That is typically 1% or so less than the total number of voters on election day, because a handful do not vote in the presidential race.

"It was the biggest election the state's ever had," said Alan Fox, an official with the firm. "It was really very overwhelming. Clerks' offices often don't have a lot of technical know-how or a lot of bodies to do all the work when it hits them."

Five communities - as small as Hales Corners and as large as Appleton - came out at 100%. Twenty-two more were above 99%, with 13 others falling within 101%, most likely because votes in the presidential race - not total ballots cast - were used as the measuring stick by the firm.

Other communities fell further out of that range. Madison and Fond du Lac, for instance, both showed 96.7% of their votes accounted for.

Neenah had a lower percentage than Milwaukee, at 95.5%. So did Mosinee (96.3%) and Eau Claire (92.5%).

The Journal Sentinel contacted some of the communities listed by the firm as having some of the largest gaps. In most cases, local election clerks reported that the actual gaps are much smaller, saying the firm got data before it was fully processed, or computer glitches in generating reports for the firm created artificial gaps.

In fact, two communities - Bayside and Waupun - said their figures reconcile exactly and officials could only speculate on why the consulting firm's figures varied.

For Waupun, the firm's analysis showed 3,948 people listed as voting, but 4,329 ballots cast in the presidential race, or 91.2% of ballots accounted for.

In Bayside, the trend was in the other direction. Bayside was listed as having 2,919 people who voted but only 2,885 ballots cast, which came out at 101.2% when measured in terms of votes accounted for.

But the actual figures in Bayside were an exact match: 2,998 voters and 2,998 ballots cast - said Village Clerk Barb Jobs. She said the consulting firm apparently failed to count the 79 people who voted in the Ozaukee County portion of the village, which would have given the firm the same number of voters as recorded by the village.

Other communities discovered gaps they didn't know existed.

In the Village of Howard, Deputy Clerk Michelle Olman found a gap of more than 200 in her figures after responding to a request from the Journal Sentinel. She said her figures show that 8,942 people voted but only 8,729 ballots were cast.

Olman said she then discovered a problem with the village's computer system. When some newly registered voters were entered into the voter database, the system automatically recorded that person as having voted in the November election, she said.

In Mosinee, the actual gap was much smaller than what was recorded by the firm, said City Clerk Bruce Jamroz.

The consulting firm said records provided by Mosinee have 2,157 voters listed, while 2,239 ballots were cast for president, putting the community at 96.3%.

But in reality, the gap was only four - 2,247 people voted and 2,251 ballots overall were cast, Jamroz said.

Jamroz and other clerks said they are constantly adding and correcting data from the November election, so the figures given to the consulting firm at any point in recent weeks are likely to differ from what the final figures show.

In Eau Claire, the review found that the city listed 34,720 voters in the November election, but 37,525 ballots were cast, putting the city at 92.5% of the votes accounted for.

Carol Schumacher, the city's election specialist, said the city is still trying to reconcile its figures, but that it has narrowed the gap reported by the consulting firm. Problems occurred because of an estimated 8,000 to 10,000 voters who changed their addresses within the city since the 2000 presidential election.

Schumacher said that when city employees enter data for new voters in the city's voter database, they are automatically prompted to designate whether those voters should be given credit for casting a ballot Nov. 2. But she said the system does not do such prompts when data is entered for voters who changed addresses within the city, which means employees must remember to give the voter credit for voting.