Friday, November 07, 2014

Numbers Part 3: Shop Talk About Turnout

Months before Election Day, I knew I'd see this:

Wrong. Wrong wrong wrong wrong wrong wrong WRONG.



Let's talk about Johnson County turnout. It's literally my job to project Johnson County turnout.

Any turnout comparison of Johnson from 2010 to 2014 needs to take into account the 2010 bar referendum. You had to be here to get what 2010 was like. Without getting too deep into my obsessive thoughts about the issue itself, it was Officially about how old you had to be to stay late in the bars. But functionally, it was a townie vs. student culture war about, as Mayor Matt Hayek said at the time, "taking our city back." For both the natives and the students, 2010 was 21 BAR!!!1! and oh yeah governor and some other unimportant stuff.

(How Johnson County is different: In 2010 I never saw one of Bob Vander Plaats' NO! signs on the Supreme Court justices until I saw a picture of one in the Des Moines Register after the election.)

A bar owner/student campaign petitioned for five solid weeks of satellite voting at every feasible dorm and classroom building, and drove president-plus level student turnout. Over 1300 people voted in one eight hour shift at the Burge dorm, which has to be a state record for a satellite site.

I worked some of those sites and despite a long ballot, no one ever spent a long time in the booth. I heard the same question over and over: "Do I have to vote on everything, or can I just vote for one thing if I want?"

Among the 17,468 early voters in Iowa City proper (almost all the student vote was at those satellite sites), there was a 9.7% undervote rate for governor, unheard of in a top of the ticket race. The undervote rate for the bottom of the ballot 21 Bar issue was just 3.7%.

The numbers were even more dramatic in the core student areas. Here's Iowa City precinct 5, which with the exception of Sally Mason is made up entirely of dorms, student apartments and Greek houses.
Total early voters:  1225
Total early votes on bar issue: 1210, 98.8% of voters (and 93% Yes for lowering the bar admission age back to 19)
Undervote 1.2%

Total early votes for governor: 813, 66.4% of total voters
Undervote 33.6%

Total straight ticket early votes: 347
Percentage of  total early vote: 28.5%
Percentage of total governor vote: 42.7%
So a third of those Iowa City 5 early voters walked into the booth and made just ONE MARK on the ballot, and another quarter of them made two marks.

County-wide (and remember, the issue was only on the ballot in the Iowa City limits), Johnson County had 2130 undervotes for governor in 2010. That's a 4% undervote rate for a top of the ticket race.

Like I said, it's part of my pay me money 8 to 5 job to project turnout, so we can plan ahead, order enough ballots, and hire enough workers. And when I did that this summer, I took into account the fact that the bar issue wasn't going to be there.

My best estimate, based on the 2010 campus satellite vote, is that about 5000 extra voters were drawn into the booth by the 21 Bar issue who would not otherwise have voted, and about half, 2500 of them, voted only on the bar issue. Another 2500 people voted because of 21 Bar, but while they were there they voted on at least some other items.

2006 was the last "normal" governor year election here, and it was at the time a turnout record, with 44,292 voters. 2006 was also a hot year, with an open seat governor race, an unpopular sixth year president, and, though I didn't know it at projection time, a wave building.

So I replaced the 2010 student early voting numbers with the same figures from 2006, which dropped my projection by about 4500. I then accounted for reprecincting and for a steady long term trend away from Election Day voting and toward early voting. (As I expected, this year was our first governor-year election with more early votes than Election Day votes. We hit that mark in the last two presidentials but fell just short in 2010.)

By August it was already clear that the absentee share of the total vote would spike due to Democratic field efforts. I also anticipated that with Terry Branstad running, the Republicans would step up their vote by mail effort as well. Branstad always had good absentee efforts in the 1990s. But in the post-Florida decade, Republicans downplayed absentees and pushed their "72 hour plan" of steering people to Election Day. When Branstad returned in 2010, he dusted off the old playbook and revived the vote by mail program.

I also assumed that population and registration would grow from 2010 to 2014 at the same rate it had between 2008 and 2012.

After plugging all that into a giant spreadsheet I came up with a turnout estimate of 51,903, split about evenly between the polls and absentees. That was down from the 53,855 total voters in 2010. But it was above the 51,725 total votes for governor.

I tweaked the numbers as I went. My last estimate on October 26 pointed to 53,433, and we adjusted the office plans from that. Better safe than sorry. But in my gut I felt like the heavy early vote would cut into Election Day turnout, so in public I stuck with the 51,903 projection.

I aimed a little low. My final projection was 24,986 at the polls, and we ended up at 25,754 (not counting 608 provisional voters, who I weigh heavier when calculating poll worker workload because they're more work). We monitored turnout through the day, and sent extra ballots out to about half the precincts just to be safe. (In the end, based on my original projections we would have run out of ballots two places: Solon and North Liberty 4.)

My calculations also pointed to 28,447 absentee ballots counted. Here I was a bit high, mainly because we had more absentees that didn't get returned than in past years. The current final absentee count is 27,185, 27,205 including 603 provisionals and late arrivals.

That's a current final total of 52,939, 52,959 with very few ballots left to add. (The last few will be counted Monday; we got only five good ones in today's mail.) Still below that 53,855 record from 2010.

But if you want to compare turnout from 2014 to 2010, and you want to compare real, non 21 Bar turnout, look at the votes, not the voters. Then you'll see that turnout in the top-ticket races is actually up slightly.  We have 52,223 votes for governor, and 52,427 in the more competitive Senate race, up from the 51,725 votes for governor in 2010.

Democrats were down three points in the governor's race, true, but Democrats were down statewide. Chet Culver had his problems but he was still an incumbent with an incumbent's name ID, while Jack Hatch was never really able to introduce himself.  The Senate races aren't good comparisons either. Even though Roxanne Conlin carried Johnson County - the only county EVER won by ANY Grassley challenger - the race was never truly competitive like Ernst-Braley looked till the very end.

The best comparison is the congressional race. It's unusual to even get a comparison this good. Same Democratic incumbent. Same Republican challenger who had run before. Same bad national climate for Democrats.
2010
Loebsack (D) 31,844  - 62%
Miller-Meeks (R) 17,920 - 35%
Sicard (L) 978 - 2%
Tack (C) 418 - 1%
Write-Ins 74

2014
Loebsack (D) 35,822 - 69%
Miller-Meeks (R) 16,231 - 31%
Write-Ins 99
Even with fewer total voters in 2014, Loebsack gained 4000 votes. And even without a Libertarian and a Constitution Party candidate eating into her share, Miller-Meeks slipped by 1700.

That's not a portrait of a "depressed Dem electorate." No, the Johnson County turnout drop is entirely explained by the absence of the polarizing local ballot issue.

No comments: