|
| EXCLUDED POLLS: |
| MORNING CONSULT - While Morning Consult produces more polls than anyone else, I do not have access to their cross-tabs and sample data. The cost to access this is prohibitive ($1500). The point of this aggregate is to increase reliability based on merging data from every pollster into a single voter sample based on 2020 exit polls. I need a poll's cross-tabs and sample data to determine if it comports with 2020 exit polls, or if it diverges, to adjust it so that it does. |
| BLOOMBERG - Bloomberg does not provide sample and cross-tab data, a practice typically utilized by those wanting to influence, not reflect, public opinion. |
| COOK POLITICAL REPORT - I only include polls with fewer than 500 respondents if it is a poll that specializes in a particular state. |
| LATEST NATIONAL POLLS |
| POLL RESULT SUMMARY (SEE FULL RESULTS & EXPLANATIONS BELOW) |
| Poll & Average Error Over Last Two Presidential Elections |
Date |
Sample + ID Bias Adjustments |
Harris |
Trump |
Pollster Avg |
Ref's Avg |
| TIPP TRACKING DAY 2 |
10/11-10/14 |
1195LV |
49 |
46 |
x |
|
| Political Ref's Take |
Neither voter sample or cross-tabs provided. Excluded. |
Excluded |
|
x |
| HARVARD-HARRIS |
10/11-10/13 |
2596LV, Inferred Party ID: 38D/26I/36R. Gender ID: 47/53 M/F. Inferred Education ID: 41/59 BA+/NoBA. Racial ID: 65/14/13/4/4 W/B/L/A/O. |
49 |
48 |
X |
|
| Political Ref's Take |
Poll Bias Score 45 towards Dems, insignificant bias. There is pro-Harris bias in three of six categories, party, gender and racial IDs. Information on two of the categories, region and age ID, is not provided. This poll barely missed the threshhold to adjust. I suspect if age had been provided, I likely would have had to adjust based on how the vote ratios tend to lean Harris. I give pollsters the benefit of the doubt in these situations, included. See calculations here. |
49 |
48 |
|
x |
| NBC NEWS |
10/4-10/8 |
1000RV, Party ID 41D/17I/42R. Gender ID: 48/52 M/F. Gender ID: Education ID: 38/62 BA+/NoBA. Racial ID: 72/12/3/2/4 W/B/L/A/O. Age ID: 16/25/33/25 18-30-45-65+. |
48 |
48 |
x |
|
| Political Ref's Take |
Poll Bias Score -133 towards GOP, severe bias, will adjust. Slightly too many GOP, too few college grads/post-grads, far too many white ID and far too few Latino ID. These pro-Trump biases are not offset by slightly pro-Harris age distribution. After adjustment. Harris gets 48.6 to Trump 47.6. See calculations here. |
48.6 |
47.6 |
|
x |
| ABC NEWS / IPSOS |
10/4-10/8 |
RV |
49.0 |
47.0 |
X |
|
| Political Ref's Take |
ABC provided neither sample or cross-tab data. Excluded for lack of transparency. |
Excluded |
|
|
| CBS NEWS / YOUGOV |
10/910/11 |
2719LV, Inferred Party ID: 38D/26I/36R. Inferred Gender ID: 47/53 M/F. Inferred Education ID: 38/62 BA+/NoBA. Inferred Racial ID: 68/12/12/3.5/3.5 W/B/L/A/O. Age ID: 16/24/34/26 18-30-45-65+. |
51 |
48 |
x |
|
| Political Ref's Take |
Poll Bias Score -38 towards GOP, insignificant bias, included in my averages. I will note that without a regional ID, it is difficult to conclude that this is actually a fair poll sample. |
51 |
48 |
|
x |
| RMG RESEARCH |
10/7-10/10 |
2945LV, Party ID: 39D/21I/40R. Gender ID: 48/52 M/F. Racial ID: 72/11/10/6 W/B/L/O. Age ID: 23/16/14/20/27 18-35-45-55-65+. |
50 |
47 |
x |
|
| Political Ref's Take |
Poll Bias Score -20, insignificant bias, included in my average. Slightly too many GOP and white ID offset by disproportionate Harris leaning age groups. Included. |
50 |
47 |
|
x |
| ECONOMIST / YOUGOV (3.7 dem) |
10/6-10/7 |
1414RV, Party ID: 33D/36I/31R. Inferred Gender ID: 46/54 M/F. Inferred Racial ID: 65/15/15/4 W/B/L/A. Inferred Age ID: 17/23/38/22 18-30-45-65+. |
47 |
44 |
x |
|
| Political Ref's Take |
Poll Bias Score 60 toward Dems. Definite bias, will Adjust. Too many Dems, too many women/few men, too few white ID and too much non-white ID. After adjusting, Harris 47.0 to Trump 44.5. |
47.0 |
44.5 |
|
x |
| PEW RESEARCH |
9/30-10/6 |
4025RV, Inferred Party ID: 37D/27I/36R. Inferred Gender ID: 48/52 M/F. Inferred Education ID: 39/61 BA+/NoBA. Inferred Age ID: 17/32/30/22 18-30-50-65+. Inferred Racial ID: 72/12/14/4 W/B/L/A. |
48 |
47 |
x |
|
| Political Ref's Take |
Poll Bias Score -35, insignificant bias, included in my average. Slightly too many Dems offset by slightly too many non-college grads and too many white ID. Overall a balanced sample, include. |
48 |
47 |
|
x |
| RASMUSSEN REPORTS |
10/3-10/9 |
2244LV, Party ID: 35D/32I/33R. Gender ID: 48/52 M/F. Racial ID: 67/13/20 W/B/O. 29/46/25 18-40-65+. |
46 |
48 |
x |
|
| Political Ref's Take |
Poll Bias Score 5, no bias, included my average. Too few GOP and too few non-white ID offset by disproportionately pro-Harris age groups. Balances well, included. |
46 |
48 |
|
x |
| REUTERS / IPSOS (+1.7 dem) |
10/4-10/7 |
969LV, Inferred Party ID: 37D/26I/36R. No other data is provided. |
47 |
45 |
x |
|
| Political Ref's Take |
Poll Bias Score cannot be determined because the pollster provided insufficient data to determine one. When a pollster does not provide the composition of its sample, one cannot determine its reliability. Excluded from my average. |
Excluded |
|
|
| NYT / SIENNA - (+8.5 dem 2020) |
9/29-10/6 |
3385LV, Party ID: 35D/32I/33R. Gender ID: 47/52 M/F. Education ID: 40/59 BA+/NoBA. Regional ID: 19/21/37/23 NE/MW/S/W. Racial ID: 65/9/11/5 W/B/L/O. Age ID: 12/22/32/28 18-30-45-65+. |
49 |
46 |
x |
|
| Political Ref's Take |
Poll Bias Score -45, insignificant bias, included in my average. The sample includes too many Dems and slightly too few men, biases offset by slightly too few non-white ID and somewhat disproportionate pro-Trump age groups. Overall the poll balances well, included in my average. |
49 |
46 |
|
x |
| CYGNAL |
10/2-10/3 |
1500LV, Party ID: 48D/7I/45R. Gender ID: 46/54 M/F. Education ID: 42/58 BA+/NoBA. Age ID: 16/23/17/19/26 18-30-45-55-65+. Racial ID: 69/12/9/3/4 W/B/L/A/O. |
47 |
45 |
x |
|
| Political Ref's Take |
Poll Bias Score -5, no bias, included in my average. Too many Dems, women and college grads/post-grads offset by too many white ID and Trump favoring age groups. The biases offset very well, balanced sample. Included in my average. |
47 |
45 |
|
x |
| YAHOO / YOUGOV (3.7 dem) |
10/2-10/4 |
1033LV, Party ID: 34D/35I/31R. Gender ID: 49/51 M/F. Education ID: 34/66 BA+/NoBA. Age ID: 21/25/33/21 18-30-45-65+. Racial ID: 63/12/16/9 W/B/L/Other. |
47 |
47 |
x |
|
| Political Ref's Take |
Poll Bias Score 40 towards Dems, or insignificant bias. Polls too many Democrats, too few white ID along with too many non-white IDs and substantially too many Dem leaning age groups. These Dem biases were offset by including too few women/many men and substantially too few college and post-grad voters. These biases offset to leave the poll with a slight Dem bias. The sample balances sufficiently, included in my average. |
47 |
47 |
|
x |
| EMERSON (+2.2 gop) |
9/29-10/1 |
1000LV, Party ID: 37D/28I/35R. Gender ID: 48/52 M/F. Education ID: 38/62 BA+/NoBA. Racial ID: 65/13/15/4/3 W/B/L/A/O. Age ID: 16/16/16/17/35 18-30-40-50-60+. Region ID: 17/38/22/23 NE/S/MW/W. |
50 |
49 |
x |
|
| Political Ref's Take |
Poll Bias Score -15, insignficant bias, included in my average. Slightly too few Republicans offset by too few college and post-grad degrees and very slightly too high southern vote. See calculations here. |
50 |
49 |
|
x |
| NPR/PBS/ MARIST (+4.0 dem) |
9/27-10/1 |
1294LV, Party ID: 37D/29I/34R. Gender ID: 49/51 M/F. Education ID: 40/60 BA+/NoBA. Regional ID: 18/23/37/23 NE/MW/S/W. Racial ID: 67/9/13/11 W/B/L/O. Age ID: 15/25/27/33 18-30-45-60+. |
50 |
48 |
x |
|
| Political Ref's Take |
Poll Bias Score 10, insignificant bias, included in my average. Too many Democrats offset by slightly too few women/many men, slightly too few college and post-grads, slightly too many south regional ID. The sample balances well, included. |
50 |
48 |
|
x |
| NEW YORK POST / LEGER (+3.5 dem 2020) |
9/27-9/29 |
851LV, Party ID: 34D/29I/29R. Gender ID: 49/51 M/F. Region ID: 18/21/38/24 NE/MW/S/W. Age ID: 30/32/38 18-35-55+. |
51 |
47 |
x |
|
| Political Ref's Take |
Poll Bias Score 185 toward Dems, severe bias. Far too few GOP and far too few southern voters, not sufficiently offset by slightly too many men. After biased areas are adjusted and averaged, Harris's lead shrinks from 4.0 to 1.7 points, Harris gets 49.7 and Trump 48.0 See calculations here. |
49.7 |
48.0 |
|
x |
| QUINNIPIAC UNIV (+5.7 dem) |
9/19-9/22 |
1728LV, Party ID: 32D/38I/30R. Gender ID: 47/53 M/F. Education ID: 52/48 BA+/NoBa. INFERRED Age ID: 11/29/44/16 18-35-50-65+ |
48 |
48 |
x |
|
| Political Ref's Take |
Poll Bias Score 185 towards Dems. Too few GOP and too many Dems, far too many college grad/post-grads and far too few non-grads. Because Quinnipiac does not provide direct ratios of each group ID, they must be inferred. Will adjust biased areas and average. See work here. After adjustments, Harris receives 47.5 to Trump 48.8. |
47.5 |
48.8 |
|
x |
| CNN / SSRS (+7.5 dem 2020) |
9/19-9/22 |
2074LV, Party ID: 32D/34I/34R. Gender ID: 48/52 M/F. Inferred Education ID: 46/54 BA+/No BA. Inferred Racial ID: 70/6/16/4 W/B/L/A. Inferred Age ID: 16/22/38/24 18>35>50>65+ |
48 |
47 |
x |
|
| Political Ref's Take |
Poll bias score 49 to Dems. Just below the threshold to maintain mild bias status, which do not require adjustment, included. I had to infer some of these IDs and they look strange. In the end, they balance. Included. |
48 |
47 |
|
x |
| OUTDATED OR DUPLICATED (2 POLLS BY SAME POLLSTER FOR DIFFERENT CLIENTS) POLLS BELOW. NOT INCLUDED IN THE AVERAGES. |
| TIPP TRACKING DAY 1 |
10/11-10/13 |
1212LV |
49 |
46 |
x |
|
| Political Ref's Take |
Neither voter sample or cross-tabs provided. Excluded. |
Excluded |
|
x |
| I&I/TIPP (2.2 dem) |
10/2-10/4 |
997LV, Inferred Party ID: 42D/17I/41R. The pollster offered no crosstabs or other data sufficient to infer the voter IDs needed to assess the poll's bias score. (gender, race, age, education). |
48 |
45 |
x |
|
| Political Ref's Take |
Insufficient data to provide a Poll Bias Score, no way to determine the composition of the sample. According to Mark Mitchell, TIPP now exclusively uses a panel to collect its data. I tried to confirm this but TIPP provides precious little information about its sample composition. But one need not scrutinize the poll sample method to know the opinion of the I&I staff on their own poll. They spend several paragraphs undermining their own findings, particularly the Independent vote of 52-36 for Harris. They point to the constant evidence of a hidden Trump vote as revealed by the "who would your neighbor vote for" question, which consistently finds Trump winning. In addition to methodolical problems, this poll provides only one of six of the voter ID measures I use to determine a poll's bias score. I cannot infer even two of the five ID categories. I&I also strongly suggests they do not believe in their own poll, preferring to interpret it's findings as proof Trump is strong despite Harris leading by three points. For these reasons I must exclude this poll. The point of my average is to get an accurate read on the current state of the race and even the publisher of this poll clearly believes that this poll can't contribute to that effort and offers no proof of its sample legitimacy. Excluded. |
Excluded |
|
|
| RMG (+2.5 dem) |
9/30-10/3 |
2965LV, Party ID: 39D/21I/40R. Gender ID: 48/52 M/F. Age ID: 21/16/15/21/28 18-35-45-55-65+. Racial ID: 72/11/11/6 W/B/L/O. |
49 |
49 |
x |
|
| Political Ref's Take |
Poll Bias Score -110 toward GOP, will adjust. Sample included slightly too few Dems, too many white ID and too few non-white IDs, disproportionate share of pro-Trump age groups (54+) over pro-Harris age groups (under 55). After adjustments, Harris receives 49.5 to Trump 48.2. See calculations here. |
49.5 |
48.2 |
|
x |
| RASMUSSEN REPORTS (1.7 gop) |
9/26-10/2 |
1762LV, Party ID: 34D/32I/33R. Gender ID: 48/52 M/F. Racial ID: 67/13/19. Age ID: 29/46/25 18-40-65+. |
47 |
49 |
x |
|
| Political Ref's Take |
Poll Bias Score -20, no bias, included in my average. Rasmussen does a daily tracker which is why the length of the fieldwork is a week. That is the usual timeframe. The New York Times usually does fieldwork for three to four days. |
47 |
49 |
|
x |
| NYT/Sienna (+8.5 dem 2020) - link |
9/11-9/16 |
2437LV, Party ID: 30D/30I/29R. Education ID: 37/63 Degree/No Degree. Racial ID: 63/12/12/3 White/Black/ Latino/Asian. Age ID: 16/23/31/25 18-29/30-44/45-64/65+. |
47 |
47 |
x |
|
| Political Ref's Take |
This poll receives a 33 lean Dem bias score, but is a mild bias and is acceptable. There is a slight education/degree bias for the GOP but an age bias for the Dems that somewhat cancel out, but not quite. Included |
47 |
47 |
|
x |
| Fox News (+2.2 dem) |
9/13-9/16 |
876 LV, Party ID: 35D/30I/35R. Gender ID: 48/52 M/F. |
50 |
48 |
x |
|
| Political Ref's Take |
This poll received a bias score of 6, but there were only two measurables, party and gender ID. When bias is uncertain, I will defer to size and quality of the poll and if the poll is an outlier. This is not an outlier and the quality is high, although in the last presidential election, Fox News polls leaned Democratic. Included. |
50 |
48 |
|
x |
| GWU/ HarrisX (+0.5 gop) |
9/14-9/16 |
1505RV, INFERRED Party ID: 38D/24I/38R. Gender ID: 48/52 M/F. Education ID: 40/60 Degree/No Degree. 68/13/13/5 White/Black/ Latino/Asian. INFERRED Age ID: 18/25/35/23 18-29/30-44/ 45-64/65+. |
51 |
49 |
x |
|
| Political Ref's Take |
Overall, this is a good sample. This poll has a bias score of -20 lean to GOP, a mild bias that is acceptible. The poll has slightly too many Republicans and slightly too few 4-year college degrees+, which is offset by too many 18-29 and 65+ and too few 30-64. Included |
51 |
49 |
|
x |
| Atlas Intel (+0.5 dem 2020) |
9/11-9/12 |
1775, Party ID: 32.4D/34.1I/33.5R. Gender ID: 47/53 M/F. Education ID: 43/57 Degree+/No Degree. Racial ID: 69/12/12/3 W/B/H/A. Age ID: 17/26/36/22 18-29/30-44/45-64/65+. |
48 |
51 |
x |
|
| Political Ref's Take |
This is an EXCELLENT sample. This poll has a bias score of 7 lean dem, which is essentially no bias at all and is included in the average at double weight. This is the most transparent poll I have seen this year. All necesary ID groups are reported. I trust this poll to actually tell me something about where the electorate stands. |
48 |
51 |
|
x |
| RMG RESEARCH (+2.5 dem) |
9/23-9/25 |
2952LV, Party ID: 46D/9I/45R. Gender ID: 47/53 M/F. Racial ID: 71/12/11/6 W/B/L/Other. Age ID: 25/16/15/19/25 18-35-45-55-65+. |
50 |
48 |
x |
|
| Political Ref's Take |
Poll Bias Score -5, no bias. Included. |
50 |
48 |
|
x |
| RASMUSSEN REPORTS (+1.7 gop) |
9/19-9/25 |
1820LV, Party ID: 35D/32I/33R. Gender ID: 48/52 M/F. Racial ID: 67/13/20 W/B/Other. Age ID: 29/46/25 18-40-65+. |
46 |
48 |
x |
|
| Political Ref's Take |
Poll Bias Score 0, or no bias. There were offsetting very slight biases in three categories, but a well balanced poll. Included |
46 |
48 |
|
x |
| NBC NEWS |
9/13-9/17 |
1000RV |
49 |
44 |
x |
|
| Political Ref's Take |
Easy take here, why are they polling all registered voters six weeks before an election instead of likely voters? Easy answer, to get a better result for Harris. As Nate Silver recognized a decade ago, " What about registered voter polls? We can infer that, because likely voter polls have no long-term bias and registered voter polls show more favorable results for Democrats, registered voter polls usually have a Democratic bias." Excluded |
Excluded |
|
|
| CBS/ YouGov |
9/18-9/20 |
3121LV, Inferred Party ID: 34D/34I/32R. Gender ID: 48/52 M/F. Inferred Education ID: 42/58 BA+/No BA. Inferred Racial ID: 72/10/10/8 W/B/L/Other. Inferred Age ID: 16/22/38/24 18>30>45>65+. |
52 |
48 |
x |
|
| Political Ref's Take - I removed this because YouGov has a bad track record over the last two presidential cycles |
Poll Bias Score 78 lean Dem. Definite and excessive bias. Correcting for party and education bias, assumptions can be seen on the bias worksheet. Also assuming Asians actually show up and vote, unlike YouGov :\ Assuming 37D/26I/37R, 40/60 BA+/No BA. Adjusted bias score is -15 lean GOP, a mild bias and is acceptable. Harris 51, Trump 49. |
51 |
49 |
|
x |
| ABC/ Ipsos - I removed this because the sample is so flawed. |
9/11-9/13 |
Party ID: 29D/30I/29R. Gender ID: 48/52 M/F. INFERRED Education ID: Ranging between 45/55 to 55/45 BA+/No BA.INFERRED Racial ID: 69/13/13/4 white/black/ hisp/asian. |
52 |
46 |
x |
|
| Political Ref's Take |
*See Row Below |
48.8 |
47.5 |
|
x |
| *Ths is a bad sample and has a lean Dem bias ranging between 52 to 172, or between an "excessive" and a "severe" bias. I will adjust to the most likely result, a midpoint between those bias levels. The sample is fine on party ID, racial ID and gender ID. It has a possible inferred education ID of 55% college degree holders to 45% non-college degree holders. This is an inferred number (educated guess) and is likely not the exact number. In a nutshell, ABC does not report crosstabs for their poll but selectivley reports a handful of data. They reported the non-college white vote, which came close to the corresponding number in the recent NYT/Sienna poll. Ipsos was 36-62 Harris-Trump, NYT was 30-66, a less favorable number for Trump. Keeping in mind that NYT found a tied race, I applied the NYT crosstabs for education, college white vote, non-college nonwhite and college non-white vote, to the ABC/Ipsos poll to determine a likely education ID range used by Ipsos. The NYT B.A.+ education vote was 61-35 Harris and the non-B.A. was 56-38 Trump. To cover my bases, I also applied a vote more favorable to Harris than NYT's to provide a range of possible education ID ratios. To see an explanation of the inference on education ID, see the first Education ID calculator here. I adjusted using the inferred education vote of degree holders voting 61-35 Harris, non-degrees 56-38 Trump using the more likely 61/39 split, and Trump actually leads Harris 47.8 to 47.0. I then applied the more favorable to Harris education vote of 64-35 Harris BA+ and 55-42 Trump non-BA, resulting in an adjusted horserace of 50.6-47.2 Harris-Trump. I will take the average of both extremes, putting the race at 48.8-47.5 Harris-Trump for my own average. The reason the education makeup of the sample is so important is because it shapes the horserace number for all other IDs in the poll, such as racial, gender and party, effectively making each ID more Trump leaning than the poll finds. When you get a sample that is too educated, all your groups are more pro-Harris regardless of their party, race or gender. Using a pool that is too educated is like dropping a lean Democrat bomb on all groups in the sample. It's a trick too many Dem leaning pollsters employ. This is why they don't release the crosstabs, because that would undermine the propaganda power of the poll. |
| Rasmussen Reports |
9/12-9/18 |
1855LV, Party ID: 35D/32I/33R. Gender ID: 47/53 M/F |
47 |
49 |
x |
|
| Political Ref's Take |
Overall, this is a good sample. This poll has a bias score of 46 toward Dems, which is not a significant bias and is acceptable. The sample is slightly too Democrat, female and educated, biases offset by being slightly too white. Included. |
47 |
49 |
|
x |
| RMG Research |
9/16-9/19 |
2969LV, Party ID: 38D/22I/40R. Gender ID: 47/53 F/M. Racial ID: 70/12/11/6 White/Black/ Hisp/Other. Age: 24/16/15/21/23 18-34/35-44/ 45-54/55-64/ 65+. |
50 |
48 |
x |
|
| Political Ref's Take |
Overall this is a good sample. This poll has a bias of -25 lean GOP, a mild bias that is acceptible. It has slightly too many Republicans and whites offset by slightly too many women and 65+. Included |
50 |
48 |
|
x |
| Yahoo/ YouGov |
9/11-9/13 |
1090LV, Party ID: 38D/24I/36R. Education ID: NOT REPORTED. Gender ID: 48/52 M/F. Racial ID: 71/12/10 W/B/H. Age ID: 15.4/23/35/27 18-29/30-44/ 45-64/65+. |
50 |
46 |
x |
|
| Political Ref's Take |
This poll has a bias score of 4.5, which is a very balanced poll. Eduation ID, however, is not reported and could represent a significant bias, but for likely voters the ratio of 4-year degrees+ to non-degrees is blank. I will include this poll, but relectantly. |
50 |
46 |
|
x |
| REUTERS / IPSOS |
9/11-9/12 |
1405RV, Neither Party ID or Education ID are provided |
47 |
42 |
x |
|
| Political Ref's Take |
This poll provides no internals to judge it's accuracy. Excluded |
Excluded |
|
|
| MORNING CONSULT |
9/11 |
3317LV, Neither Party ID or Education ID are provided affordably |
50 |
45 |
x |
|
| Political Ref's Take |
This poll provides no internals to judge it's accuracy for a reasonable price. Excluded |
Excluded |
|
|
| RASMUSSEN REPORTS |
9/5-9/11 |
1838LV, Inferred Party ID: 35D/32I/32R Gender Makeup: 52/48 F/M. Racial Makeup: 67/13/20 White/Black/Other |
47 |
49 |
x |
|
| Political Ref's Take |
Party Vote: D's 82-16 Harris, I's 51-39 Trump, R's 82-16 Trump. I think the D+3 is too high and the I's are too high. Reallocating the party vote to the more likely turnout of 37D/26I/36R, Harris gets 46.2 to Trump 48.7. |
46.2 |
48.7 |
|
x |
| NYT / Sienna |
9/3-9/6 |
1695LV, Party ID: 31D/28I/33R. Education ID: 41% Degree, 59% No Degree. |
47 |
48 |
x |
|
| Political Ref's Take |
*See Row Below |
46.7 |
47.5 |
|
x |
| *Party Vote: D 91-4 Harris, I 46-41 Harris, R 94-5 Trump. Education Vote: Degree 57-38 Harris, No Degree 56-39 Trump. Using more likely Party ID 37D/26I/36R, Harris gets 47.4 to Trump 46.0. Using more likley education ID, Harris 46.0 to Trump 49.0. Averaging party and education IDs together, Harris 46.7 to Trump 47.5. In summary, they included too many Republicans but also too many college degrees. The oversamples offset each other more or less. |
| Harvard- Harris |
9/4-9/5 |
1695LV, Party ID: D37/23I/36R. Education ID: 66/34 Degree/No Degree |
50 |
50 |
x |
|
| Political Ref's Take |
The party and education IDs are good. Included |
50 |
50 |
|
x |
| NPR / Marist |
9/3-9/5 |
1413RV, Party ID: 38D/26I/35R, Education ID: 61/39 Degree/ No Degree |
49 |
48 |
x |
|
| Political Ref's Take |
Party Vote: D 94-6 Harris, I 49-46 Trump, 94-5 Trump. The education ID is reasonable. Applying the party vote to more likely party ID of 37D/26I/36R, Harris gets 48.5 and Trump 48.8. |
48.5 |
48.8 |
|
x |
| RMG RESEARCH (Scott Rasmussen) |
9/3-9/5 |
2701LV, 40D/7LeanD/6I/ 7LeanR/40R |
50 |
48 |
x |
|
| Political Ref's Take |
*See row below |
50 |
48 |
|
x |
| *Party Vote: D's 95-4 Harris, Lean D 92-6 Harris, I 45-40 Harris, Lean R 88-6 Trump, R 92-7 Trump. Education turnout not reported, but reports how each education group voted. Grad degree +21 Harris, 4-year degree Harris +6, no degree Trump +4. This is a smaller than normal lead for Trump with the no degree group. Nonetheless, the sample and weighting look decent, Included. |
| EMERSON COLLEGE |
9/3-9/4 |
*See row below |
49 |
47 |
x |
|
| *1000LV, 37D/28I/35R, 2020 Vote 49-45 Biden (+4.3), Education: 38% college degree, 62% no college degree. Region: 17% Northeast, 38% South, 22% Midwest, 23% West. |
| Political Ref's Take |
*See row below |
48.8 |
47.0 |
|
x |
| *Overall the only internal that sticks out is party turnout. I do not think this electorate will be 1% less Repulbican than 2020. Adjusting to more likely 37D/26I/36R. Party Vote: D's 90-7Harris, I's 50-42 Harris, R's 93-7 Trump. Reallocating that vote to the more likely voter turnout, Harris 48.8 to Trump 47.0. |
| ECONOMIST / YOUGOV |
9/1-9/3 |
1389RV, 33D/36I/31R |
47 |
45 |
x |
|
| Political Ref's Take |
*See row below |
47.8 |
47.5 |
|
x |
| *Party vote: D's 92-2 Harris, I's 36-31 Trump, R's 91-5 Trump. The poll reports that it weights for racial makeup, age, education, but NOT region. We have no idea if they oversampled higher educated voters, which is common. Or if they included too many Northeasterners, etc. Not reporting this data makes it impossible to fully trust this poll. Will reallocate the party vote to the more likely 37D/26I/36R turnout. Also, because this is a registered voter poll, will allocate the I column to a value of 1 to eliminate unlikely voters. Adjusted party vote: D's 92-2 Harris, I's 54-46 Trump, R's 91-5 Trump. Harris 47.8 to Trump 47.5. |
| IBD/TIPP |
8/28-8/30 |
1386RV, Inferred party ID 37D/26I/36R |
48 |
45 |
x |
|
| Political Ref's Take |
*See row below |
46.5 |
47.0 |
|
x |
| *The problem with this sample is that it includes all registered voters, which almost always includes slightly too many Democratic leaning voters in each party (R/I/D). Party Vote Among Registered: D's 92-5 Harris, I's 43-42 Trump, R's 88-9 Trump. Adjusting the party vote to 90-7 Harris, 46-41 Trump, 90-7 Trump. Harris 46.5 to Trump 47. Also suggesting this is a pro-Trump sample, Trump wins the "who would your neighbors vote for" by 43-39, a key question that indicates a shy Trump vote. |
| WSJ |
8/24-8/28 |
1500RV, Party ID: 35D/23I/35R; Education: 39 Degree, 61 No Degree; Region: East 18%, South 39%, Midwest 22%, West 22% |
48 |
47 |
x |
|
| Political Ref's Take |
This is a registered voter poll, which typically favors Dems marginally. The sample has slightly too few Dems, which arguably offsets the party ID ratio, so no adjustment necessary. Included |
48 |
47 |
|
x |
| Qunnipiac Univ |
8/23-8/27 |
1611LV, 32D/38I/30R |
49 |
48 |
x |
|
| Political Ref's Take |
Party Vote: D's 98-1 Harris, I's 45-45 Tie, R's 94-5 Trump. Something funky is going on in this sample. They don't report racial makeup, education or region, so I don't know who they've polled too much. The D+2 isn't great, but this poll is just hard to adjust. Will leave as is, included. |
49 |
48 |
|
x |
| YAHOO / YOUGOV |
8/22-8/26 |
1194RV, Party ID: 33D/26I/29R; Education: 66% No Degree, 34% Degree; Region: not provided; Racial makeup: not provided. |
47 |
46 |
x |
|
| Political Ref's Take |
Party Vote: D's 96-2 Harris, I's 44-35 Trump, R's 92-2 Trump. Obviously this registered vote poll includes a lot of people who won't turn out, so the I column will be allocated to a 1 value. New party vote is D's 96-2 Harris, I's 55-45 Trump, R's 92-2 Trump. Reallocating vote to the more likely 37D/26I/36R, Harris 47.9 to Trump 48.2. |
47.9 |
48.2 |
|
x |
| ABC/Ipsos |
8/23-8/27 |
2496 Adults, 29D/30I/29R |
50 |
46 |
x |
|
| Political Ref's Take |
The internals of this poll lack key indicators, like number of registered or likely voters polled, party ID of registered and likely voters, how the parties voted, education, region or racial makeup. It is misleading to publish results without making access to your internals reasonably available to the general public. Excluded. |
Exc- luded |
|
|
|
| MORNING CONSULT |
9/2-9/4 |
11414LV |
49 |
46 |
x |
|
| Political Ref's Take |
To see internals for MorningConsult costs $1500 per year. It is misleading to publish results without making access to your internals reasonably available to the general public. Excluded. |
Exc- luded |
|
|
|
| RASMUSSEN REPORTS |
8/29-9/4 |
1838LV, Inferred Party ID: 35D/32I/32R; Racial Makeup: |
46 |
47 |
x |
|
| Political Ref's Take |
Party Vote: D's 84-12 Harris, I's 46-41 Trump, R's 87-11 Trump. I think the D+3 is too high and the I's are too high. Reallocating the party vote to the more likely turnout of 37D/26I/36R, Harris 45.7 to Trump 47.7. |
45.7 |
47.7 |
|
x |
| |
| POLITICAL REF'S CURRENT TURNOUT MODEL: Party ID: 37D/26I/37R. Gender ID: 48/52 M/F. Education ID: 60% no degree, 40% College Degree. Racial ID: 67% White, 13% Black, 13% Latino, 4% Asian, 4% Other. Regional ID: 20% East, 23% Midwest, 35% South, 22% West. I think the party split will be 50/50 because both Gallup and Pew have found that more Americans identify as Republican or Republican leaning in massive surveys this year. I believe this is actually a conservative estimate, when it may be more likely that Republicans and Republican leaners could outnumber their Democrat couterparts based on the Gallup and Pew numbers. |
| Back to the top of the page |
| TURNOUT MODEL EXPLAINED: I apply a turnout model to every poll and determine a bias score. I give considerable deference to pollster modeling. Once the bias is definite and excessive, however, I will correct the poll before including it in the Political Ref's average. The pollster's topline is always included in the pollster average as is. I look at five key poll sample categories, party ID, gender ID, education ID, racial ID and age ID. If any one of these categories is missing, the poll's credibility suffers. |
| Back to the top of the page |
| WHAT POLLS DO I INCLUDE? I mostly include polls that RealClearPolitics.com will include, although I occassionally add a poll that I think brings good data. I disagree with some of their decisions about what polls to include, but becuase most fair minded political observers will look at RCP I want to offer a take on the average. |
| Back to the top of the page |
Political Ref's Poll Average