Keith Ellison, the Representative from Minnesota’s Fifth Congressional District and candidate for Democratic National Committee chair, has cited turnout improvement in his Minneapolis-based district as one of his main qualifications for the position.
Just this past Saturday, for example, Ellison’s campaign account Tweeted this graphic:
In an interview with Vox’s Ezra Klein, Ellison went into more detail, making a number of specific claims (my emphasis):
When I was elected in 2006, my district had the lowest turnout in the state of Minnesota. Now it’s the highest, and it’s consistently the highest. One of the reasons why is because we focus on turnout 365 days of year. […] When I first got into office, I had 150,000 votes. Now if we don’t get 250,000 votes, we feel disappointed. Because we got 250,000 this year and we got 262,000 in 2012. […] In 2014, voter turnout statewide decreased in the state of Minnesota from 2010, hitting a 70-year low. In my district, we increased turnout by 3 percent in 2014. Even in down years, we’re going up. It’s the only congressional district in Minnesota, the fifth district, where voter turnout grew between 2010 and 2014. That’s what we’re doing, and that’s why I think I need to be the DNC chair.
Klein even echoed the point: “Those are very impressive numbers, and particularly the point about 2014 turnout increasing in your district. That’s an achievement.” These claims are what give Ellison’s detailed descriptions of his operations — “the canvassing, the door knocking, the calling” — the weight of authority. I imagine they give the impression of “here’s someone who knows how to get people to the polls”.
I can’t speak to Representative Ellison’s motivations, but by all accounts he’s a sharp politician. Presumably he knows how much a lot of partisans, and maybe especially the Democratic party officials who vote on DNC chair, like the idea of winning elections by “organizing” to increase turnout (you only have to appeal to people you agree with!). Anecdotally, the narrative of “here’s someone who knows how to get people to the polls” has reached a fairly wide audience.
Unfortunately, while I have no personal opinion about whether he’d be a good DNC chair or not, most of Ellison’s numbers are misleading. He makes comparisons across years with different boundaries, uses absolute totals instead of percentages, and doesn’t mention how his own initial vote total was depressed by a third-party candidate. When more conventional metrics are used, there’s little evidence of increasing turnout in Ellison’s districts, and in fact turnout may be decreasing.
Keith Ellison was elected to Congress in 2006. Here are the total votes cast by Congressional District in Minnesota in each general election from 2004 to 2016, with dotted lines across the 2010 reapportionment. Ellison’s district is highlighted.
Ellison is right about at least one thing: In 2006, his district did have the state’s lowest “turnout” in absolute terms, casting only 250,544 votes. But he doesn’t mention that it still had the state’s lowest “turnout” in absolute terms in 2010 (just 234,323 votes) after his first two terms. Both of these are related to something else he doesn’t mention: From 2006 to 2011, MN-05 also had the state’s smallest or second-smallest estimated citizen voting-age population (CVAP) in annual ACS estimates. If you have fewer people in your district, then, all else being equal, you’ll also have fewer voters.
The absolute “turnout” in MN-05 really only increased after 2010—that is, after reapportionment, when the district was redrawn to equalize its population. Because MN-05 had too few residents under the old lines, redistricting added over 40,000 people in initial counts in Brooklyn Center and parts of Edina and Hopkins, and its organic population growth may have increased as well. (As far as I can tell the district boundaries didn’t shrink anywhere but I could be wrong.) Ellison’s boast that his district cast all of 6,386 more votes in 2014 than 2010 seems a bit less like it’s about a “very impressive” “achievement” in that context.
Nor is he correct that MN-05 was the “only congressional district in Minnesota…where voter turnout grew between 2010 and 2014”: The St. Paul-based MN-04, which also added population in redistricting, cast 9,659 more votes in 2014 than in 2010. Actually, MN-05 is now one of the largest districts in the state by population or by CVAP, and even so it still isn’t “the highest, and consistently the highest” turnout district as Ellison claimed above, not even in absolute terms.
A more conventional measure of “turnout” anyway is in percentage terms, not absolute terms: total votes divided by CVAP, which can be (perhaps noisily) interpolated from the Census’ American Community Survey estimates. I’m not sure if I used the standard way to interpolate CVAP, so take it with some caveats (Minneapolis claimed 70% turnout of registered voters in 2004 but this approach gives them 76% turnout by CVAP — I wonder if they were double-counting voters who re-registered in person, though). With that in mind, by this standard, turnout in MN-05 might actually be trending down:
Looking at midterm cycles, turnout in the old MN-05 went from 59.7% of CVAP in 2006 to 52.3% of CVAP in 2010 and then (after redistricting) fell further to 47.6% of CVAP in 2014. Looking at Presidential cycles, the turnout percentage dropped from 81.0% to 79.9% in 2004 and 2008 (in the old MN-05) and then from 76.4% to 75.1% in 2012 and 2016 (in the new MN-05).
Ellison has been saying his district went from the lowest-turnout to the highest-turnout district. The actual highest-turnout district, in every one of these elections, has been MN-05’s neighbor, the affluent and educated MN-03. Even following Ellison and comparing across redistricting, MN-05 was third in 2004 and 2006, second in 2008, last in 2010, fourth in 2012, second-to-last in 2014, and fourth again in 2016. Frankly, that seems a lot more consistent with the national pattern of those elections than Ellison’s account.
Even if you look at, for example, DFL votes in topline races only (President in Presidential cycles, Governor in midterm cycles) — which isn’t the commonly-accepted meaning of “district turnout” anyway — Ellison’s claims fare little better, either in absolute terms:
Or in percentage terms:
Finally: In both his campaign’s chart and in his Vox interview, Ellison uses his initial 2006 election to Congress as a baseline (“When I first got into office, I had 150,000 votes. Now if we don’t get 250,000 votes, we feel disappointed”). This is a pretty poor choice of baseline. In fact Ellison only got 136,060 votes in his initial election in 2006 — probably because a Minnesota Independence Candidate, Tammy Lee, got 21%, nearly beating the Republican.
Ellison’s predecessor, Martin Sabo, had won with 171,572 votes in the previous (Republican-friendly) midterm of 2002, although the district may have shrunk some in between. If Ellison and his team “feel disappointed” when they don’t get 250,000 votes, they must have been pretty disappointed as recently as 2014, when he won a larger district with just 167,079 votes, not to mention in 2010, when he only won with 154,833 votes. See the OurCampaigns history here:
This isn’t a full analysis of Ellison’s turnout record. For that you’d have to compare Minneapolis, MN-05, or Hennepin County to similar places (like well-educated urban areas in states that are somewhat contested but usually not among the very most targeted), as well as analyze what his actual operations have been. For all I know they’ve been effective by some metric. But the case that Ellison has been making about his own turnout record is pretty shaky, and probably shouldn’t have been so widely disseminated in its current form.
Notes and Sources:
Vote totals by Congressional District are rolled up from Minnesota’s excellent precinct election results spreadsheets for 2004–2016. Note that at least in 2008 there were a handful of (decisive!) votes in the statewide recount which do not appear to be officially assigned to any precinct. CVAP estimates are interpolated for November 1st for each election year from the American Community Survey’s B05003 tables for 2006–2015, using 1-year estimates, except for 2006 when only one estimate was available, and taking them as estimates for July 1st of their listed year — IE, midyear (as these estimates aren’t for any particular date) . This interpolation method was inspired by looking at Elect Project but the particulars are my choices. The nearest two years that were within the same redistricting period around November 1st were used. Estimates for 2005 were available, but looked like they were too low — using them would have increased turnout for 2004, making my point stronger! I have checked this simple interpolation math repeatedly but there might always be some mistakes.
