A few factoids about elections in Waukesha County, WI in 2004 and 2006 are getting more play than, at this point, they appear to deserve. It doesn't appear that too many people voted in 2004, or that there were actually more votes than ballots in 2006. It's not impossible, but it is far from obvious.
Summary meta: I'm not trying to shut down the investigation in Waukesha County (how could I do that, anyway?); I'm not trying to defend Kathy Nickolaus's record as county clerk; and I'm not trying to reassure people about elections in general. No matter what happened in Waukesha County last decade, the country has serious election problems. If you think it's objectively evil or counterproductive to try to figure out whether specific claims hold up, we simply disagree about that.
(An earlier diary presents some of my thoughts about the 2011 controversy.)
Kathy Nickolaus was elected county clerk in 2002.
November 2004 turnout
The issue: The November 2004 summary report on the county website shows 231,031 ballots cast and 236,642 registered voters, for an eye-popping 97.63% voter turnout. Somewhere, Nickolaus actually brags about that turnout. It's ridiculous on its face.
Discussion: Yes, it is, and at best Nickolaus should be embarrassed that she bragged about it. Check out some other numbers:
- Here is the September 2004 county summary, which also shows 236,642 registered voters. That number is pretty much impossible. Wisconsin has Election Day Registration, so if the September figure is approximately correct, then the November figure should be substantially higher.
- Here is the February 15, 2005 summary report, which shows 282,390 registered voters. That's a difference of almost 46,000 registered voters, or about 20%. A huge registration drive after the presidential election? Probably not. Probably most of those new registrations came on or before election day.
- Can we find any other numbers to support that inference? I haven't yet spotted any figures for election day registrations on election day 2004. In 2006, the state reported 3,012,525 early registrations, 43,851 late registrations, and 392,391 Election Day registrations (see page 103) -- so the late and election day registrations provided a 14+% bump. (Waukesha Co.'s 2006 "bump" was smaller, about 9%.) Wisconsin was a battleground state in 2004, so one would expect a substantially larger bump in that year than two years later.
- But isn't 282,390 registered voters an awful lot for a county that size? I'll update this diary with specific numbers, but the short answer is: not so much. Registration rolls generally include some people who have died or left the county (or who moved within the county and are temporarily double-registered). That's a big problem if the dead people "vote," otherwise mostly a nuisance. Waukesha Co.'s registration rate in 2004 (based on the February 2005 numbers) is similar to Dane Co.'s.
- Anyway, we started off with suspiciously high turnout in 2004, so let's take a moment to ask: did suspiciously many people vote in Waukesha Co. in 2004, regardless of how many people were registered? According to Dave Leip's Atlas of Presidential Elections, in 2004, the total vote in Waukesha was 230,363 out of 2,997,007 statewide, or 7.69%. In 2000 (before Nickolaus arrived), the figures were 203,734 out of 2,598,607, or 7.84%. So the vote count in 2004 doesn't seem anomalously high. (Nate Silver did similar calculations for more recent elections.)
None of this information establishes how many people really were registered in Waukesha Co. in 2004, and how many voted. But if we assume that the registration number in the November summary should have been close to the figure reported in February 2005, the numbers generally fall into place.
November 2006 ballots cast
The issue: The Waukesha Co. November 2006 summary report shows 156,804 ballots cast, but a total of 176,112 votes in the gubernatorial contest -- a difference of 19,308 votes. How could there be more votes than ballots cast?
(Could I just mention? That report also shows 210 precincts counted out of 211. Why? I don't know, although I have a pretty good idea. Has any previous diary mentioned that? Not to my knowledge. Just saying.)
Discussion: Here is what Nickolaus says now:
Ballot Cast is the number of ballots that were fed through the election machines at the polling places and the results were collected using a modem in the office. It does NOT include any hand entered results.
I don't want to take her word for it, and I'm sure you don't either. Does it make any sense? Here's something that the
Milwaukee Journal Sentinel apparently reported after the September primary(!) (
Reddit source):
Because Menomonee Falls and Mukwonago use a different brand of equipment than the rest of the county, those two municipalities were unable to file Tuesday's returns electronically with the county. That, in turn, contributed to a slower-than-usual reporting of countywide results....
Nickolaus said she has new software that will combine results from the incompatible equipment, and she hopes to have that issue resolved by November. On Tuesday, however, she did not want to post incomplete results on the Web site that excluded Menomonee Falls and Mukwonago.
OK, that could be a clue.
Now, check out the state's Fall 2006 Voter Turnout and Registration report -- Waukesha Co. is on pp. 95-96. It shows 177,424 voters, including 17,265 in Menomonee Falls and 2,591 in Mukwonago -- i.e., 19,856 voters in the two villages that were interoperability fails in the September primary.
So, if we suppose that Kathy Nickolaus's programming wizardry was not equal to the task of solving the reporting problem between September and November (cue Vizzini...), and that the votes from these two villages were in fact entered by hand, that would more than account for the gap between ballots cast and votes counted.
Say, what about that missing precinct? The canvass report provides a clue (this is a snippet from the gubernatorial contest):
0199 V MERTON W1 2 & 3 258 898 6 0
0202 V MUKWONAGO W1 - 8 891 1651 33 3
0203 V MUKWONAGO W10 0 0 0 0
0205 V NASHOTAH W1 & 2 220 491 6 0
0208 V NORTH PRAIRIE W1 2 & 3 308 637 15 0
Mukwonago ward 10 is missing. Oops. That could be an interesting lead, or just another reporting error. Or it's possible that the ward was actually phased out. I'll just say that comparing the numbers in the county website reports with the numbers in the state report probably won't give you a strong sense of confidence that every vote was counted and reported accurately. The phrase "close enough for government work" might spring to mind. And, yes, that is snark.
However, I don't see much reason to think that about 20,000 votes were "stuffed," either. In the state's 2002 canvass report, Waukesha Co. reported 142,114 votes out of 1,775,349 statewide, or 8.00%. In the 2006 state report, Waukesha's 177,424 reported voters out of 2,183,155 voters statewide make 8.13%. (Note that both percentages are higher than in the presidential elections.) Take away 20,000 votes in 2006, and you'd be around 7.2%. Facially, it doesn't seem very likely.
The other thing that doesn't seem likely to me is that Nickolaus somehow cleverly added about 20,000 votes without any of the municipal clerks or canvass board members noticing -- but wasn't clever enough to change the ballots cast figure.
Conclusion
In the final analysis, did Kathy Nickolaus report correct vote counts in 2004 and 2006 (and 2011 and...)? How the heck should I know? The vote counts could be wrong in any number of ways. And even if we could convince ourselves that none of Nickolaus's mistakes and, umm, quirks has ever materially affected an election outcome -- hey, it's a big country. There is a lot to fix. If worrying about Waukesha helps us do election verification better, that's great. If it turns into a massive distraction, not so great.