For most of July we’ve seen polling from average and mediocre (and sometimes terrible) pollsters. That was not surprising. The last wave of polling from top line polling outfits came at the end of June and I did not expect that to resume until after the conventions. However, the conventions are very late this year. This week, we saw the return of three excellent pollsters: Marist (A+); Monmouth (A+), and; Fox News (A-).
In July, we saw the presidential race slightly tighten. I think that is probably because of the number of new polls from low-rated pollsters who often have the race closer (Rasmussen [C+]; Zogby Analytics [C+]; Harris [C]). There were some good pollsters in the field before this week, but they were mostly B-rated polls. Monmouth did do some polling early in House districts (which I hope they are continuing to do), but it was not until this week that the A-rated pollsters have gotten back into the picture after the last polling wave in June.
Let’s look at the different results we are seeing between A-rated pollsters and C-rated pollsters. Today, Marist and Rasmussen have polls out. Marist is not just one of six A+ rated pollsters from FiveThirtyEight, but it is arguably the best of those six. Marist’s final poll in 2016 was the closest to the actual result of the popular vote (within one point). Considering the bad rap polling got out of that election, Marist can claim top dog until proven differently. Rasmussen is overrated in my opinion and probably deserves a D rating. That’s because I think they engage in push-polling (asking questions designed to frame public opinion rather than report it) and applying too-generous weighting protocols – particularly in turnout modeling – that favor Republicans early in campaign cycles. Then, as the election grows closer they seem to start being more honest so their final polls are much closer to reality than earlier polls were.
In today’s surveys, both pollsters included presidential approval questions. Marist polled 1,118 registered voters and found that just 39% of voters approve of Trump’s performance as president (representing a -18 point disadvantage for Trump). Rasmussen, however, found 47% approving for a net -5 points. They claim to have surveyed 1,500 likely voters. I am using less certain language about Rasmussen’s methods because – like so many other poor pollsters – they are not transparent about it. Marist is transparent to a fault, as pollsters should be. Here’s another thing that should make you doubt Rasmussen: In July and early August, they several times found Trump to have a net positive approval rating, including getting approval rating as high as 50% and 51%. A lot of pollsters are surveying approval ratings. Rasmussen’s findings are not credible not just in comparison with Marist, but in comparison with almost every other pollster. So, you say, Rasmussen might know something the rest don’t. Nope. They don’t.
This example of Marist vs. Rasmussen demonstrates what’s been largely been missing in the polling since June. Lower-rated (and just plain poor) pollsters have released data that has Trump doing better than have higher-rated (and much better) pollsters. Polling averages – particularly FiveThirtyEight’s (Real Clear Politics is not consistent about this) – tend to weight more heavily more recent polls. It’s true that there is weighting related to the rating of the pollsters too, but the sheer number of polling from lower-rated pollsters, which is not balanced by higher-rated pollsters (which during July were mostly B-rated pollsters) can contribute to a tightening of the race that might not be true. So, at the beginning of July – just after the June polling wave – FiveThirtyEight polling average was Biden +9.6. By the first week of August it was Biden +7.6. This week it is up to Biden +8.1.
Let’s take a look at this week’s polls. For the national presidential polls, let’s look at this week’s results by pollster rating.
A-rated pollsters
Marist (A+): Biden 53 – Trump 42; Biden +11
Monmouth (A+): Biden 51 – Trump 41; Biden+10
Fox News (A-): Biden 49 -Trump 42; Biden +7
Georgetown University/Battleground (A/B): Biden 53 – Trump 40; Biden +13
Beacon Research/Shaw & Co. (A-): Biden 48 – Trump 41; Biden +7
Simple polling average: Biden +9.6
B-rated pollsters
Data for Progress (B-): Biden 52 – Trump 40; Biden +13
Morning Consult (B/C): Biden 51 – Trump 43; Biden +8
Pew Research Center (B/C): Biden 53 – Trump 45; Biden +8
Ipsos (B-): Biden 58 – Trump 42: Biden +16
YouGov (B): Boden 49 – Trump 39; Biden +10
RMG Research (B/C) Biden 45 – Trump 37; Biden +8
Simple polling average: Biden +10.5
C-rated and below pollsters
HarrisX (C): Biden 44 – Trump 40; Biden +4
Rasmussen Reports (C+): Biden 49 – Trump 43; Biden +6
Change Research (C-): Biden 50 – Trump 44; Biden +6
Simple polling average: Biden +5.3
As you can see the A and B rated pollsters have Biden with a much more commanding lead than the C-rated pollsters do. The A and B pollsters are finding similar results, but the C-rated pollsters have Biden’s margin half that of the higher rated pollsters. I think we will soon see some other A+ rated pollsters – such as ABC News/Washington Post and Siena College – resume polling nationally. The Democratic National Convention begins next week with the GOP convention the following. If history is any indication (and it might not be this year since preferences are so strong in this election cycle already), Biden and then Trump will see some sort of bump in support during their convention weeks and perhaps immediately afterwards. I would have expected the best pollsters to wait until after the conventions to resume their surveys, but as I noted above the conventions are late this year. Clearly, Marist and Monmouth were not willing to wait that long before getting polls back in the field. Perhaps other will join them this weekend (I don’t see much utility polling during the convention weeks, but some will do it).
Nothing has changed this week on the consensus electoral map. Biden is forecast to win 320 electoral votes and the Toss-Up states remain Texas, Iowa, Ohio, Georgia, and North Carolina.
Senate Races
A number of Senate races were surveyed this week. Let’s break it down by state.
Arizona: Change Research (C-) has Mark Kelly leading by six points over GOP Sen. Martha McSally (49-43). Emerson College (A-) has Kelly up by 11 points (52-41). OH Predictive Insights (B/C) has Kelly up by five (48-43).
Georgia: In the regular election, HIT Strategies (no rating) has Democrat Jon Ossof leading GOP Sen. David Perdue by three points (42-39. SurveyUSA (A) has Perdue up by three points (44-41). In the special election, MRG Research (B/C) has GOP Rep. Dough Collins leading the jungle primary by four points with 27% of the vote. In second place is Democrat Raphael Warnock with 23%, followed by incumbent GOP Sen. Kelly Loeffler at 21%. HIT Strategies finds Loeffler up by four points with 22% of the vote, followed by Collins with 18%. SurveyUSA has Loeffler up by nine points with 26% of the vote, followed by Collins and Warnock each with 17%.
Kansas: It’s starting to look like nominating Rep. Roger Marshall has not created the slam dunk in the race that establishment Republicans were hoping for. SurveyUSA polled over 1,200 likely voters and found Marshall up by just two points over Democrat Barbara Bollier (46-44). This is starting to look like a very competitive race.
Maine: Critical Insights (C/D) has state Speaker Sara Gideon leading GOP Sen. Susan Collins by five points among likely voters (43-38) and eight points among registered voters (43-35). RMG Research (B/C) has Gideon up by seven points (48-41).
Michigan: Change Research (C-) has Democratic Sen. Gary Peters leading Republican John James by three points (48-45).
Minnesota: Emerson College (A-) has Democratic Sen. Tina Smith leading Republican Jason Lewis by three points (48-45). There has not been a lot of polling on this race, but earlier polls had Smith leading by 9-15 points. If this poll is not an outlier, it suggests that this race might be a lot more competitive than anticipated.
Mississippi: This race looks closer than anticipated in a poll this week by Garin-Hart-Yang Research Group (B/C). GOP Sen. Cindy Hyde-Smith is leading Democrat Mike Espy by five points (47-42). Three percent of respondents chose Libertarian candidate Jimmy Edwards. This is important because if no one gets 50% on November 3rd, there will be a run-off between Hyde-Smith and Espy on November 24th. Like other states that have not been considered competitive, there has been little polling in Mississippi. But, in the polls we have there does seem to be a trend of the race narrowing since earlier this year.
North Carolina: Harper Polling (B/C) has Democrat Cal Cunningham leading GOP Sen. Thom Tillis by three points (41-38). If Tillis – as an incumbent US Senator – is really at 38% support as we get towards the end of August then he’s in big trouble. Emerson College also has Cunningham up by three (44-41). Change Research has Cunningham up by five points (48-43). Public Policy Polling (B) has Cunningham leading by four points (48-44).
South Carolina: Public Policy Polling finds GOP Sen. Lindsey Graham leading Democrat Jaime Harrison by three points (47-44). Graham has a serious challenge on his hand in the Palmetto State.
The Cook Political Report is going on record this week forecasting as much as a net +10 seat pick-up for the Democrats in the House. Hopefully, by Labor Day Monmouth and others will have a lot more district polling and we can get a sense whether that supports CPR’s forecast.
That’s the polling run-down for this week. We’ll do it again next week. Also for next week, look for the first trifecta state update (it will be North Carolina) and a piece on Sen. Kamala Harris (D-CA) and how Vice Presidential nominees help or hurt the ticket.