-

Corwin D. Smidt (smidtc@msu.edu)

Headline numbers

  • Voterfile-based predicted turnout rate among registered voters: Depending on assumptions, ranges in between 53-54%.

  • Expected number of voters: Depending on assumptions, ranges in between 4.35-4.43 million voters.

  • Adding uncertainty, it has been hard to measure uncertainty with changes in the law. The estimate has been off 1-3 percentage points in past years. So that suggests a reasonable range of outcomes is 50-57 percentage points or 4.1-4.7 million.

So turnout rates look very much likely to be lower than 2018 (58%), and higher than other recent midterms (such as 2014 and 2010). A comparable election would be 2006 (53%). But, considering the larger number of registered voters, we would still expect over 4 million votes. This expectation is consistent with the pattern we saw in August 2022. Not as many people voted in 2022 as in 2020, but it is still one of only three August elections that saw over 2 million people vote.

What can we learn about Michigan’s electorate?

The continued purpling of Michigan’s electorate

Just based on registration patterns alone Michigan is becoming more a more closely divided state.

Despite voterfile-based expectations of relatively high midterm turnout (which helped them in 2006), the balance of voters across the state continues to shift away from Democratic strongholds. Political scientists find little evidence that higher turnout benefits Democrats over Republicans. This would hold for 2022 and Michigan.

We continue to see consistent declines in the relative number of registered voters in Democratic counties. This includes in Wayne County, and new signs of decline in Oakland. In contrast, we see continued growth in Macomb, Kent, and Ottawa.

So even if everyone votes for the same party as they have over the last ten years, we would expect an improvement among Republicans. To show this, I set each county’s expected vote to be equal to the average presidential vote margin (Democrat %-Republican %)

The Turnout Gap Among Congressional Districts

Further evidence of the GOP advantage in turnout can be seen by comparing turnout expectations in Congressional districts. These were redrawn last year to be nearly equal in population. But a comparison of turnout rates shows there is a larger difference in the size of each district’s electorate, most notably the two Detroit districts (12 and 13).

District Party Lean Exp. # of Voters
1 Rep 388481
2 Rep 337486
3 Even 328768
4 Rep 331418
5 Rep 315404
6 Dem 373517
7 Even 355469
8 Even 338296
9 Rep 369400
10 Rep 334764
11 Dem 370109
12 Dem 303275
13 Dem 257529

2022 Turnout by Congressional District

Michigan voters will be older, but the youth vote is showing staying power

Compared to 2018, the expectation is that this midterm electorate will be older; the median age among voters is likely 57. The median age in 2018 was 55, which was much closer to the lower ages we see in presidential elections (53 in 2016 and 52 in 2020).

Although this change in age favors Republicans, the trends show it isn’t all good news for Republicans. Compare the relative size of voting groups under 60 years old.

Age Group 2014 2016 2018 2020 2022
18-29 6.0 13.6 11.7 15.5 ~11.9
30-44 16.8 21.3 20.2 21.9 ~18.6
45-59 30.6 29.1 27.8 26.0 ~25.2
60+ 46.6 36.1 40.3 36.6 ~44.3
Median 58 53 55 52 ~57

The size of the youth vote is expected to increase over 2018. That is the most pro-Democratic group in the state. In comparison advanced middle age voters (45-59) continue to decline in relative size. In Michigan polls, this Generation X group of 45-59 year-olds leans the most towards the GOP. This change in relative size is largely driven by the increased number of registrations among these younger groups. For example, a breakdown of registration numbers and turnout rates show, that the 45-59 group is now a smaller portion of registered voters in Michigan than the 30-44 year old group. If these voters stay in Michigan, we should expect this group of younger voters to become much more influential as the baby boomer generation declines in Michigan.

Age Group Millions Registered Exp. Turnout Rate Millions Voting
18-29 1.44 35% 0.53
30-44 2.03 41% 0.82
45-59 1.97 56% 1.11
60+ 2.74 72% 1.95

Expected County Turnout

Number County Exp. # of Voters
1 ALCONA 5819
2 ALGER 4578
3 ALLEGAN 54577
4 ALPENA 14186
5 ANTRIM 14206
6 ARENAC 7294
7 BARAGA 2983
8 BARRY 30677
9 BAY 48272
10 BENZIE 10797
11 BERRIEN 64104
12 BRANCH 17129
13 CALHOUN 50963
14 CASS 21495
15 CHARLEVOIX 14645
16 CHEBOYGAN 12550
17 CHIPPEWA 14053
18 CLARE 13099
19 CLINTON 39006
20 CRAWFORD 6852
21 DELTA 17106
22 DICKINSON 10977
23 EATON 50178
24 EMMET 19755
25 GENESEE 171797
26 GLADWIN 11721
27 GOGEBIC 6693
28 GD. TRAVERSE 51066
29 GRATIOT 15894
30 HILLSDALE 18618
31 HOUGHTON 14553
32 HURON 14734
33 INGHAM 113329
34 IONIA 25763
35 IOSCO 13182
36 IRON 5962
37 ISABELLA 21871
38 JACKSON 63400
39 KALAMAZOO 112315
40 KALKASKA 8922
41 KENT 287448
42 KEWEENAW 1371
43 LAKE 5408
44 LAPEER 42948
45 LEELANAU 15007
46 LENAWEE 39664
47 LIVINGSTON 103821
48 LUCE 3054
49 MACKINAC 6159
50 MACOMB 385038
51 MANISTEE 12407
52 MARQUETTE 30246
53 MASON 15010
54 MECOSTA 16901
55 MENOMINEE 9683
56 MIDLAND 41136
57 MISSAUKEE 7038
58 MONROE 65483
59 MONTCALM 27236
60 MONTMORENCY 5378
61 MUSKEGON 71618
62 NEWAYGO 22219
63 OAKLAND 624368
64 OCEANA 12002
65 OGEMAW 10400
66 ONTONAGON 3284
67 OSCEOLA 9853
68 OSCODA 4259
69 OTSEGO 11872
70 OTTAWA 138362
71 PRESQUE ISLE 7083
72 ROSCOMMON 12936
73 SAGINAW 81917
74 ST. CLAIR 17079
75 ST. JOSEPH 3281
76 SANILAC 32435
77 SCHOOLCRAFT 71878
78 SHIAWASSEE 27324
79 TUSCOLA 21753
80 VAN BUREN 31598
81 WASHTENAW 179199
82 WAYNE 678457
83 WEXFORD 15123

2022 Turnout by County

Context and background: the Michigan Qualified Voter File (QVF) and voter registration laws

The QVF is a state records of registered voters and their recent turnout history. The Secretary of State is mandated by law to keep a centralized, active list of qualified electors in Michigan along with their recent voting turnout histories.

But the data set is not historic record. Michigan registration records suffer inaccuracies because of our decentralized nature and mismatched laws. The Secretary of State relies on a network of 1500+ local clerks to inform the state of changes in registration status and voter turnout records. The State shares the responsibility of registering voters and administering the QVF because the state allows voter registration when applying for a drivers license, but local clerks also register voters and are responsible for keeping voter turnout records and reporting these records to the state.

And the recent constitutional amendment passed by voters, 2018’s Proposal 3, made it easier for residents to register (and vote), even if they live for a short time in Michigan. But there was no consideration of its impact on state registration requirements and the laws for removing past registrants.

For example, the state currently reports we have a total of 8.2 million registered voters. This is 10% larger than registered voters in 2018 despite little population growth. And this total is larger than scholarly estimates of Michigan’s Census voting-eligible population, which is 7.9 million.

A primary reason for this mismatch is that state law requires a lengthy process for the identification and cancellation of voter registrations. The recent state QVF I received this month, currently lists the total number of voters slated for cancellation at over 670,000 (8%).

The Fall 2020 QVF listed around 460,000 (6%) as slated for cancellation. And about 11-12% of those slated voters return to active status by voting in subsequent elections. In 2018 (before Proposal 3 passed), the rate of voters slated for cancellation in the QVF was about 5%.

Interestingly, northern counties are the most over-registered, perhaps reflecting the impact of the pandemic on choice of residency during the fall of 2020. And counties with large universities (e.g., Ingham, Issabella, Mecosta, Washtenaw) have the largest proportion of voters slated for cancellation, reflecting the transient status of college students in these areas.

Prediction track record

How these predictions are made

Model estimates of voter turnout are generated from looking at the turnout history registered voter in Michigan as listed in the state’s database September. It estimates how age, address, and past voting behavior typically predicts voter turnout. And then predicts an individual-level (and jurisdiction-level) estimate of the probability of voting for 2022, and then it simulates an election across these observations.

Model estimates allow for a year-specific tide, and this is currently estimated from voter turnout in the August statewide party primary. This year, I adjust this prediction among the 6% of voters slated for cancellation. I estimate that about only 22-26% of those could still vote in Michigan across 2018 and 2020.

Model specifics

Define y\_{ijt} as the set of QVF observations of whether a registered voter i in precinct j turned out to vote in election t. Voter probabilities of turning out to vote are then modeled as follows:

\\Pr\[y\_{ijt} = 1\] = \\Lambda( x\_{it}'\\beta + z\_{t}'\\delta
+ Aug\_t( \\gamma\_{1} + \\zeta\_{1j} + \\nu\_{1i} ) +
Competition\_t(\\gamma\_{2} + \\nu\_{2i}) + \\zeta\_{0j} + \\nu\_{0i}
)
 + Competitiont(\gamma{2} + \nu{2i}) + \zeta{0j} + \nu_{0i} ) “)
where

  • x\_{it} are of set time-varying individual level predictors: logarithm of age;the interaction of log age with August primaries; the interaction of log age with the midterm cycle; a logarithm of month since registration; and dummies for an initial observation, whether a voter had recently registered in that quarter, and whether that recent regisration was during a presidential cycle.
  • z\_{t} are a set of election-specific indicators: whether the contest is a presidential primary; whether that election is during a midterm cycle; and the year of the contest.
  • Aug\_t is an indicator of whether the election is an August primary. The relationship between this variable and a voter turnout is allowed to vary at the precinct level and at the individual-level (\\zeta\_{1j} +
\\nu\_{1i}). In other words, August turnout is assumed typically higher or lower in some precincts and among some individuals for factors beyond what is in the model. This specification essentially allows some individuals and precincts to be high turnout precincts in November but not in August (and vice-versa).
  • Competition\_t is a (logged) measure counting statewide and U.S. House contests with major party competition or, for August, with primary competition. This effect is allowed to vary by individual (\\nu\_{2i}). In short, it allows that some people show up to vote regardless of the amount of competition on the ballot. Whereas other people’s likelihood of voting is responsive to the amount of competition on the ballot.
  • \\zeta\_{0j} is a precinct-specific, time constant parameter that captures that precinct’s unexplained tendency to have higher or lower registered voter turnout rates.
  • \\nu\_{0i} is an individual-specific, time constant parameter that captures that individual’s unexplained tendency to have higher or lower voter turnout rates.

Subsampling and prediction generation

Multilevel logit model estimates were performed over 10 separate subsamples of 100 randomly sampled precincts (weighted by population) with up to 50 randomly sampled voters within each precinct. Since the average voter has a voter history spanning back ten past contests. There are about 50,000 observations for each of the 10 estimates. These parameters are then averaged and used to generate Empirical Bayes estimates of the precinct and individual specific error components (\\zeta,
\\nu), using 5-7 integration points. These Empirical Bayes estimates are then combined with the model’s fixed portion to generate a prediction of each voter’s probability of turning out to vote in November.