MICHIGAN POLLS: Peters +3.7 - Peters 49.5, James 45.8 REF'S: Peters +3.0 - Peters 49.0, James 46.0 RETURN TO SENATE RACES 2016 Exit Polls - 40D/29I/31R without leaners (PDF) 2018 Exit Polls - 42D/15I/43R with leaners (PDF) Straight Poll Averages Chart Freedom Window Averages Chart Poll Date Sample + Party Distribution Result Pollster Avg Ref's Avg Emerson College 10/29-10/31 700LV, 36D/25I/39R Peters 52, James 46 x   Political Ref's Take on the above poll Party distribution is reasonable. No adjustment necessary. Included   x RMG Research 10/27-10/29 800LV, Not provided Peters 50, James 41 x   Political Ref's Take on the above poll Strong GOP model Peters 48, James 42   x Tarrance Group 10/24-10/56 Not provided Peters 48, James 46 x   Political Ref's Take on the above poll No adjustment necessary. Included   x Trafalgar Group 10/15-10/18 1034LV, Not provided Peters 48, James 50 x   Political Ref's Take on the above poll No adjustment necessary. Included   x Trafalgar Group 9/26-9/28 1042LV, Party distribution not provided Peters 48, James 47 Repeat   Political Ref's Take on the above poll This was among the most accurate battleground state polls in 2016. Included   Repeat Marist College 9/19-9/23 799LV, 31D/36I/31R Peters 49, James 44 >1month   Political Ref's Take on the above poll Party distribution is reasonable. No adjustment necessary. Included   >1month Baldwin Wallace Univ (Great Lakes Poll) 9/8-9/22 1001LV, 41D/22I/37R Peters 46, James 41 >1month   Political Ref's Take on the above poll Party distribution is reasonable. No adjustment necessary. Included   >1month Morning Consult 9/11-9/20 1376LV, Party distribution not provided Peters 47, James 40 >1month   Political Ref's Take on the above poll Neither party distribution or voter ratios within those parties are provided. Excluded     Marketing Resource Group 9/14-9/19 600LV, +4 Dem sample Peters 42, James 40 >1month   Political Ref's Take on the above poll Party distribution is reasonable. No adjustment necessary. Included   >1month Ipsos 9/11-9/16 637LV, 45D/11I/39R Peters 49, James 43 >1month   Political Ref's Take on the above poll Party distribution is reasonable. No adjustment necessary. Included   >1month EPIC-MRA 9/10-9/15 600LV, 43D/11I/39R Peters 45, James 41 >1month   Political Ref's Take on the above poll Party distribution is reasonable. No adjustment necessary. Included   >1month Benenson Strategy Group/GS Strategy Group 8/28-9/8 1600LV, Party distribution not provided Peters 45, James 41 >1month   Political Ref's Take on the above poll Biden is strong among seniors and those over fifty in Michigan where he holds a 54-40 lead, so James is trending ahead of Trump's numbers with this group and the rest. Included   >1month Rasmussen Reports 9/2-9/3 1000LV, Party distribution not provided Peters 48, James 40 >1month   Political Ref's Take on the above poll Neither party distribution nor voter ratios of those parties is provided. Excluded     Change Research 9/4-9/6 876LV, Party distribution not provided Peters 50, James 46 >1month   Political Ref's Take on the above poll Neither party distribution nor voter ratios of those parties is provided. Excluded     Glengariff Group 9/1-9/3 600LV, Party distribution not provided Peters 44, James 41 >1month   Political Ref's Take on the above poll The poll notes that this Senate race will look very similar to the presidential race. While that is true, James holds a better favorable/unfavorable ratio than Peters and may be trending up. Included   >1month Tarrance Group 9/1-9/3 569RV, Party distribution not provided Peters 47, James 46 >1month   Political Ref's Take on the above poll Neither party distribution or voter ratios within those parties are provided. Excluded     Change Research 8/21-8/23 809LV, Party distribution not provided Peters 50, James 45 Repeat   Political Ref's Take on the above poll Neither party distribution or voter ratios within those parties are provided. Excluded     Trafalgar Group 8/14-8/23 1048LV, Party Distribution not Provided Peters 47, James 48 Repeat   Political Ref's Take on the above poll Neither party distribution or voter ratios within those parties are provided. Excluded     Tarrance Group 8/10-8/13 602RV, Party distribution not provided Peters 49, James 44 Repeat   Political Ref's Take on the above poll Neither party distribution or voter ratios within those parties are provided. Excluded     Change Research 8/7-8/9 413LV, D+4 across all battlegrounds, but party distribution for each state is not provided Peters 48, James 45 Repeat   Political Ref's Take on the above poll Neither voter turnout assumptions or voter ratios within the parties is provided. Impossible to verify credibility of the poll. Excluded     Gravis Marketing 7/22 754LV, 37D/34I/30R Peters 49, James 39 >1month   Political Ref's Take on the above poll The party distribution has too many Independents, but the D+7 is defensible. That comes close to 2016. No adjustment necessary Included   >1month Fox News 7/18-7/20 756RV, 44D/16I/40R Peters 48, James 38 >1month   Political Ref's Take on the above poll This is a voter sample that pushed leaners to choose a party. When comparing polls that push leaners to pick a party with those that do not, Michigan looks more Republican when leaners are pushed to pick. So this voter turnout assumption is consistent with the 2016 and 2018 exit polls because it is presidential year. It should not match 2018 exactly because the GOP turns out better in non-presidential years. It should fall somewhere between R+1 and D+9, which it does, at D+4. No adjustment necessary. It should be noted, however, that this is a registered voter poll which almost always favor Democrats. The margin is likely closer than what this poll shows. Included   >1month Spry Strategies 7/11-7/16 700LV, 31.4D/36I/29.6R Peters 47, James 37 >1month   Political Ref's Take on the above poll The Independent column is somewhat high, suggesting that some leaners picked a party and some did not. In 2018 when leaners were pushed to pick, the state was R+1. In 2016 when they were not forced, the state was D+9. So it's hard to say. The result between Peters and James is consistent with other polls with a solid voter turnout assumption. Bottom line, the high amount of Independents makes this one difficult, but on balance I think no adjustment is necessary. Another point, this is a likely voter poll with a seemingly strong likely voter screen. I am more reluctant to adjust polls with solid likely voter screens. They pick up on who is actually going to come out and vote. Included.   >1month PPP 7/9-7/10 1041RV, 38D/29I/33R Peters 49, James 42 >1month   Political Ref's Take on the above poll The D+5 sample undersamples Democrats. Will adjust to the more likely 2016 model, 40D/29I/31R. Voter ratios within the parties are not provided but can be inferred. Peters had a D+5 turnout assumption but a 7 point lead, which suggests he was stronger among Independents and Republicans than James was with Independents and Democrats. Will assume each candidate won their party 90-5 and that Peters won Independents by 2, or 15.5-13-5 or 53% to 47%. Reallocating these voter ratios to the more likely turnout, Peters receives 52.9 to James 43.5. Peters 52.9, James 43.5   >1month Change Research 7/10-7/12 824, Party Distribution Not available Peters 50, James 43 Repeat   Political Ref's Take on the above poll Neither party distribution or voter ratios within those parties are provided. Adjustment not possible. Excluded from our average. Adjustment not possible. Excluded from our average.     Change Research 6/26-6/28 699LV, The overall sample of all battleground states was 39D/27I/34R. Peters 49, James 41 Repeat   Political Ref's Take on the above poll The +5 D sample in battleground states is an oversample of Democrats. Adjustment is impossible because we don't know the party distribution of each state. Excluded from our averages. Excluded from our averages     PPP 6/26-6/27 1237RV, 36D/31I/33R Peters 47, James 39 Repeat   Political Ref's Take on the above poll The D+3 sample undersamples Democrats. Michigan is tricky because when you push leaners to pick a side the state looks more Republican. I do think that Michigan should end up with about 40% of voters identifying Democrats and 29% Independent and 31% Republican in terms of actual registered voters. Adjusting to 40D/29I/31R. The voter ratios within parties is not provided but can be inferred. Peters had a +3 voter turnout advantage and still won by 8. So he likely won Independents and more Republicans than James won Democrats. For calculation purposes only, I will assume each candidate won his party 90-5 and Peters won Independents 18.5-13-5 or 60% to 40%. Reallocating these voter ratios to the more likely results, Peters receives 55 to James 42. Peters 55, James 42   Repeat Marketing Resource Group 6/12-6/15 600LV, Party Distribution Unavailable Peters 36, James 30 >1month   Political Ref's Take on the above poll Neither voter turnout model or Peters/James vote ratios withing parties provided. No adjustment possible. Excluded     American Greatness/TIPP 6/9-6/12 907RV, 37.2D/28.6I/29.9R Peters 47, James 35 >1month   Political Ref's Take on the above poll The party distribution is very close to the 2016 turnout. No adjustment necessary. No adjustment necessary   >1 month Kiaer Research 5/31-6/7 543LV, 37D/35I/28R without leaners. 57D/43R with leaners. Peters 48, James 32 >1 month   Political Ref's Take on the above poll The party distribution is close to the 2016 exit poll. No adjustment necessary. No adjustment necessary   >1 month EPIC-MRA 5/30-6/3 600LV, 43D/11I/38R Peters 51, James 36 >1 month   Political Ref's Take on the above poll If anything, the sample includes slightly too many Republicans based on the 2016 exits. With the Dem showing a huge lead, no adjustment necessary. No adjustment necessary   >1 month PPP 5/29-5/30 1582RV, 39D/31I/30R Peters 48, James 39 Repeat   Political Ref's Take on the above poll Solid voter turnout assumption, matches 2016 almost exactly. No adjustment necessary. No adjustment necessary   Repeat