Monday, June 22, 2020

Massive Dem +7.62% "Trump Approval"Aggregate Polling Bias Exposed

This is, I believe, the first and only analysis of the aggregate Dem bias in polling which makes up the RCP and 538. com Trump Approval ratings. RCP doesn't include many of these not so common polls but their aggregate is also distorted by lumping all polls they carry whether Registered/Likely/All Voters polling. Likely voters polls have a significantly higher level of Trump approval than All Voters polls. There is usually a 1-2 point higher approval for Trump at RCP, currently 43% than 538 which gives a choice of viewing an AV poll aggregate or an RV/LV aggregate, currently 42.4%.

The Dem aggregate bias from the 29 polls which detail their D/R/I split is D+6.20%. Rasmussen uses D+4 as does Selzer. D+4 is, according to exhaustive research as presented by PPD's Richard Baris a realistic baseline given the California/NY/IL popular vote Dem advantage. The 2016 election actual result 
was D+2.

THE 20 POLLS ABOVE THE D+4 BASELINE HAVE A D+7.62% AGGREGATE DEM BIAS**



A partisan advantage is not the only method of bias/distortion. Females may be over-represented, rural voters may be under-represented which was the polls failing in 2016, education and income levels may be wrongly calculated and of course turnout modeling as Nguyen rightly states will have an effect. For a comprehensive review of the electorate see the Pew Analysis. That said the over-representation of Dems in 19 of 29 polls is clearly the significant factor in Trump's polling.

A further 11 polls which are included in the RCP/538 aggregates do not provide their demographics, it could well be surmised that all of them are above the D+4 baseline especially the likes of the ABC/Washington Post poll and the true aggregate is over D+7. All these polls, their D biases and methodology,can be found at the RCP/538 websites linked above. 538 arbitrarily "adjusts" Rasmussen down 4 points. It is a matter of conjecture, but if all these polls were D+4 and Likely Voters, Trump's approval would be at least 50%

 Nota Bene: The +7.65/6.20 figures were correct when this article was first written as the pollsters change their demographics it will vary but will be circa that figure-higher or lower as time goes on;

AGGREGATE DEM LEAN +6.20%  (LV=Likely Voters RV=Registered Voters AV=All Voters)

MORNING CONSULT D+5
NBC/WSJ/HART RESEARCH RV D+12
IPSOS REUTERS D+ 13
**ECONOMIST/YOUGOV D+11 or D+9
AMERICAN RESEARCH GROUP ARG RV  D+10
QUINNIPIAC   RV  D+10
CBS/SSR  AV D+9.4
SIENNA/NYTIMES RV D+9
GLOBAL STRATEGY GROUP/GBAO RV D+8
CLIMATE NEXUS RV D+7.5
KAISER FOUNDATION D+7.5
***CNBC/CHANGE RESEARCH LV D+7 
CNN/SRS RV D+7
EMERSON POLLING LV D+7% (+3.8 in June)
TARGET SMART/DYNATA RV D+7
MARIST/PBS RV D+7
GALLUP AV D+6.6
OPTIMUS D+6.5
SURVEY MONKEY AV  D+6.3
WASHINGTON POST/ABC RV D+6
FOX NEWS   RV D+6
SUFFOLK/USA TODAY RV D+5.8
HARVARD/HARRIS RV D+5
HARRISX/THE HILL RV D+5
HART RESEARCH RV D+4
SELZER/GRINNELL LV  D+4
RASMUSSEN REPORTS LV D+4
MONMOUTH  RV D+4 ("Unweighted" D+6.2%)
YOUGOV D+4 (Democrat 29% Republican 25% Independent 32%)

McLAUGHLIN & ASSOCIATES RV D+3
PPP (D) RV D+2
ROANOKE AV D+2
DEMOCRACY INSTITUTE LV; EVEN
ZOGBY POLL LV; EVEN



D/R/I INFORMATION NOT SUPPLIED (D=DEM/R=GOP/I=IND)


AXIOS/SURVEY MONKEY No DRI 
ECHELON INSIGHTS LV NO DRI
YOUGOV "Trump Tweets" No DRI 
MSR GROUP RV No DRI 
WHITMAN STRATEGIES 500 RV NO DRI
IBD/TIPP RV Poll No DRI

RMG (Scott Rasmussen) RV DRI not supplied
REDFIELD AND WINTON RV NO DRI
MEETING STREETS INSIGHTS RV NO DRI



Alex Castellanos-one of the very few to predict a Trump victory in 2016







METHODOLOGY

*Morning Consult;PID: Dem (no lean) 770 39% PID: Ind (no lean) 554 28% PID: Rep (no lean) 669 34%

"2012 Vote: Barack Obama 578 38% 2012 Vote: Mitt Romney 322 21% 2012 Vote: Other 44 3% 2012 Vote: Didn’t Vote 565 37% N 1508" 

The New York Times/Siena College Research Institute June 17-22, 2020 1,337 United States Registered Voters MOE +/- 3.0%



WASHINGTON POST/ABC

Partisan divisions are 30-24-39 percent, Democrats-Republicans-independents
Margin of error RV  4.0%



ROANOKE



DEMOCRACY INSTITUTE (2020) LV
Dem 38%  GOP 38%  Independents 24%


SURVEY MONKEY 
Party ID
AnswerTotalRepublican/ lean repIndependent no leanDemocrat/ lean dem
Unweighted N50,40419,4457,74921957

RASMUSSEN REPORTS Likely Voters D+4




KAISER FOUNDATION D+7.5%





CBS AV D+9.4


SELZER/GRINNELL LV  D+4     
Strong Republican 16 Republican 10 Lean Republican 7 Totally independent 26 Lean Democrat 10 Democrat 11 Strong Democrat  16 Other 2 (VOL) Refused/ not sure 3
1,009 U.S. adults ages 18 and over March 27-30, 2020 Including 777 likely voters in the 2020 general election Weighted by sex, age, race, Margin of error: ±3.1 percentage points based on all and educational attainment Margin of error: ±3.5 percentage points based on 2020 likely voters

AMERICAN RESEARCH GROUP (ARG) "Mother Jones observed in 2007 that ARG "may not be the bellwether for accurate polling."[5] The New Republic found in 2008 that "ARG is a black sheep of the polling world; I repeatedly heard it singled out for scorn by other pollsters."
Sample Size: 1,100 completed interviews among a random sample of all adults age 18 and older living in households in the continental United States (see methodology for details). There were 259 Republicans, 355 Democrats, 377 independents, and 109 non-voters in the sample.
Margin of Error: � 3 percentage points, 95% of the time, on questions where opinion is evenly 

HART RESEARCH D+4
1,015 registered voters nationwide
Strong Republican. 19  Leaning Republican 19 Completely independent 20 Leaning Democratic. 19 Strong Democrat 23 Total Republican 38 Total Democrat 42

CLIMATE NEXUS RV  D+7.5
Sample Population: 1917 registered voters in the United States. Sample Selection: Scientific online poll - stratified sample of panel respondents. Weighting Parameters: The sample was weighted based on the U.S. Census Bureau’s Voting and Registration Supplement to the Current Population Survey for registered voters in the United States based on age, gender, race, educational attainment, census region, and Hispanic ethnicity. This topline provides weighted percentages, as well as the unweighted N-size for the total sample. Due to the effects of weighting and rounding, figures may or may not add up to 100%. The standard deviation of the weights was: 0.2242216. The maximum weight was: 1.9601923. The minimum weight was: 0.3773268. 95% of the weights were between 0.4793321, 1.4957711. Margin of Error: The 95% credibility interval for this survey is +/- 2.3%, which includes the square root of the design effect (DEFT): 1.0248166.

PPP (D) RV D+2
Public Policy Polling surveyed 1,403 national registered voters from April 23-24, 2020 on behalf of Protect Our Care. The margin of error is +/- 2.6%. 50% of interviews for the survey were conducted by telephone and 50% by text





HARVARD/HARRIS RV D+5
Democrat 37% Republican 32%  Independent 28% Other 3%
 Monthly Harvard-Harris Poll: May 2020 This survey was conducted online within the United States from May 13-14, 2020 among 1,854 registered voters by The Harris Poll. The results reflect a nationally representative sample. Results were weighted for age within gender, region, race/ethnicity, marital status, household size, income, employment, political party, political ideology, and education where necessary to align them with their actual proportions in the population. Propensity score weighting was also used to adjust for respondents’ propensity to be online.



ABC/WASHINGTON POST RV NO DRI
Washington Post-ABC News poll was conducted by telephone among a random national sample of 1,001 adults, with 75% reached on cell phones and 25% on landlines. Results have an error margin of +/- 3.5 percentage points for the full sample including design effects due to weighting. *=0.5%

ECHELON INSIGHTS LV NO DRI
We surveyed N = 1,000 registered voters in the Likely Electorate online from May 18 - May 20, 2020 using a voter file-matched sample as part of our monthly Verified Voter Omnibus tracking survey of the 2020 political environment. The sample was weighted to known characteristics of the 2020 “Likely Electorate” population, a frame which takes into account demographic and turnout characteristics of the 2020 electorate.

TARGET SMART/DYNATA RV D+7
Democrat and Ind. Lean Dem 47 - Pure Independent 10 - Republican and Ind. Lean Republican 40
TargetSmart designed this multi-mode survey and administered it in partnership with industry-leading data collection vendor Dynata. Six hundred (556 weighted) respondents were interviewed over the telephone by professional agents (79% on cell phones, 21% on landlines), and 600 (644 weighted) respondents were interviewed online, reached via the Dynata online panel. Telephone interviews were drawn from a random sample of registered voters across the United States, sampled from the TargetSmart voter file. Online interviews from the Dynata online panel were also all matched back to the TargetSmart voter file. In total, the survey reached 1,200 respondents who indicated they are registered to vote in the United States. The survey was conducted from May 21- 27, 2020. Quotas were designed to reflect the demographic and geographic distribution of registered voters in the United States. The data were weighted by gender, age, modeled race, modeled partisanship, modeled education, modeled vote propensity, population density (at the census bloc level), and geography (at the state level) to ensure an accurate reflection of the population.

MONMOUTH D+4
DEMOGRAPHICS (weighted) Self-Reported 27% Republican 41% Independent 31% Democrat
The Monmouth University Poll was sponsored and conducted by the Monmouth University Polling Institute from May 28 to June 1, 2020 with a national random sample of 807 adults age 18 and older. This includes 279 contacted by a live interviewer on a landline telephone and 528 contacted by a live interviewer on a cell phone,

AXIOS/SURVEY MONKEY No D/R/I split provided. All Voters 2016; 23% did not vote; 16% GOP 51% Ind 19% Dem 6,275 Unweighted respondents. 2020 vote 5,666 Unweighted Trump GOP 93% Ind 35% Dem 2%


YOUGOV No DRI Uses All Voters and Registered Voters
U.S. citizens, aged 18 and over. Sampling method Respondents were selected from YouGov’s opt-in Internet panel using sample matching. A random sample (stratified by gender, age, race, education, geographic region, and voter registration) was selected from the 2016 American Community Study. Voter registration was imputed from the November 2016 Current Population Survey Registration and Voting Supplement.

CNN/SRS D+7
A total of 1,259 adults, including an oversample of 250 black, non-Hispanic adults, were interviewed by telephone nationwide by live interviewers calling both landline and cell phones. Oversampled black, non-Hispanic respondents have been weighted to represent their proper share of the adult population. Interviews were conducted in English and Spanish. Among the entire sample, 32% described themselves as Democrats, 25% described themselves as Republicans, and 44% described themselves as independents or members of another party.The study was conducted for CNN via telephone by SSRS, an independent research company. Interviews were conducted from June 2-5, 2020 among a sample of 1,259 respondents, including an oversample of 250 black, nonHispanic respondents. The landline total respondents were 448 and there were 811 cell phone respondents. The margin of sampling error for total respondents is +/- 3.4 at the 95% confidence level. The design effect is 1.47.



MSR GROUP RV No DRI
Data collected on-line using national panel sampling frame. A random sample of 1000 adults yielded a final data set of 855 registered voters. Data were collected on June 7, 2020. Results were weighted to national norms based on post-stratification by age and gender.

WHITMAN STRATEGIES 500 RV NO DRI “Whitman Insight Strategies helps Democrats and Independents motivate their Base, “

FOX NEWS 1343  RV D+6

SUFFOLK UNIVERSITY D +5.8
"Youngest member of the household"


IBD/TIPP RV Poll No DRI
IBD/TIPP Poll reflects an online survey of 1,233 adults from May 31 to June 3. The Trump-Biden 2020 presidential election poll reflects a subset of 964 registered voters. 



Zogby Poll LV Even 36/36/I 28
Zogby Analytics conducted an online survey of 1007 likely voters in the US.
Using internal and trusted interactive partner resources, thousands of adults were randomly invited to participate in this interactive survey. Each invitation is password coded and secure so that one respondent can only access the survey one time.
Using information based on census data, voter registration figures, CIA fact books and exit polls, we use complex weighting techniques to best represent the demographics of the population being surveyed. Weighted variables may include age, race, gender, region, party, education, and religion. The party breakdown for this survey is as follows: 36% Democrat, 36% Republican and 28% Independent/unaffiliated.


EMERSON POLLING RV NO DRI
National Emerson College/Nexstar Media poll was conducted June 2-3, 2020. The sample consisted of registered Democratic, Republican, and Independent voters, n=1431, with a Credibility Interval (CI) similar to a poll’s margin of error (MOE) of +/- 2.5 percentage points. The data sets were weighted by age, ethnicity, education, and party based on 2016 election turnout.

MARIST/PBS D+6 RV
This survey of 1,062 adults was conducted June 2nd through June 3rd, 2020 by The Marist Poll sponsored in partnership with NPR and PBS NewsHour. Adults 18 years of age and older residing in the contiguous United States, including an oversample of non-white Americans, were contacted on landline or mobile numbers and interviewed by telephone using live interviewers.
RV 51% Women 49% Men Party Identification Democrat 38% Republican 31% Independent 29% Respondents Dem Men 15% Women 23% GOP Men 16% GOP Women 15%

OPTIMUS RV D 6.5 “ PARTY REPUBLICAN 28.5 DEMOCRATIC 35 INDEPENDENT/OTHER 36.3”

QUINNIPIAC July 2020




** ECONOMIST/YOUGOV

I emailed them. It says they get their party affiliation numbers from 2014
533D/390R/577I. 801 female. 699 male. Adds up to 1500. Reports Registered and All Voters.
They put down 1221 was RV but the numbers add up to 1500.



McLAUGHLIN AND ASSOCIATES D+3  D36 R33 I31 1000 LV

IPSOS/REUTERS D +13



GLOBAL STRATEGY GROUP/GBAO NAVIGATOR RESEARCH D+8 RV


1,009
When it comes to politics, do you generally think of yourself as a strong Democrat, not very strong Democrat, strong Republican, not very strong Republican, an independent, or some other political party? [IF INDEPENDENT/OTHER] Do you think of yourself as closer to the Democratic Party or the Republican Party? Strong Democrat 30% Weak Democrat 12  Independent / Closer to the Democrats Independent 11  Independent / Closer to the Republicans 5 Weak Republican  13 Strong Republican 22



DEMOCRAT (NET)  43%.INDEPENDENT (NET) 22 REPUBLICAN (NET) 36 DEMOCRAT W/ LEANERS (NET) 49% REPUBLICAN W/LEANERS (NET) 41


MORNING CONSULT (POLITICO) RV D+5
 Methodology: This poll was conducted between June 12-June 14, 2020 among a national sample of 1987 Registered Voters. The interviews were conducted online and the data were weighted to approximate a target sample of Registered Voters based on gender, educational attainment, age, race, race, and region. Results from the full survey have a margin of error of plus or minus 2 percentage points.


***CNBC/CHANGE RESEARCH D+7

Was D+5%  May 2020 switched to D+7% 6/26

Generally speaking, do you think of yourself as a Democrat, a Republican, or an independent? (IF RESPONDENT SAYS "DEMOCRAT" OR "REPUBLICAN," ASK:) Would you call yourself a strong (Democrat/Republican) or not a very strong (Democrat/Republican)? (IF RESPONDENT SAYS "INDEPENDENT," ASK:) Do you think of yourself as closer to the Republican Party, closer to the Democratic Party, or do you think of yourself as strictly independent? Strong Democrat ................................................ 24 [153] Not very strong Democrat ................................... 10 Independent/closer to Democratic Party ............. 10 Strictly independent ............................................ 15 Independent/closer to Republican Party ............. 8 Not very strong Republican ................................ 10 Strong Republican .............................................. 19 Not sure ............................................................ 4 Total Democrat 44 Total Republican 37 


RMG (Scott Rasmussen) RV DRI not supplied
 The survey of 1,200 Registered Voters was conducted by Scott Rasmussen using a mixed mode approach from June 11-13, 2020. Field work for the survey was conducted by RMG Research, Inc. Most respondents were contacted online or via text while 306 were contacted using automated phone polling techniques. Certain quotas were applied, and the overall sample was lightly weighted by geography, gender, age, race, education, and political party to reasonably reflect the nation’s population of Registered Voters. Other variables were reviewed to ensure that the final sample is representative of that population.

Abacus Data RV No DRI 
The survey was conducted with 1,500 American adults from June 11 to 13, 2020. A random sample of panelists was invited to complete the survey from a set of partner panels based on the Lucid exchange platform. These partners are typically double opt-in survey panels, blended to manage out potential skews in the data from a single source.
The margin of error for a comparable probability-based random sample of the same size is +/- 2.5%, 19 times out of 20.
The data were weighted according to census data to ensure that the sample matched the United States’ population according to age, gender, race, educational attainment, and region. Totals may not add up to 100 due to rounding.

NBC/WSJ/HART RESEARCH RV

Latest Poll D+7

Previous Poll D+12
Strong Democrat 28 Not very strong Democrat 7 Independent/lean Democrat 10 Strictly Independent 13 Independent/lean Republican 7 Not very strong Republican  6 Strong Republican 20 Other (VOL) 5 Not sure 4 + Results shown reflect responses among registered voters
1000 Registered Voters, including 550 respondents with a cell phone only and Date: May 28-June 2, 2020 11 respondents reached on a cell phone but who also have a landline.

REDFIELD AND WINTON RV NO DRI
Sampling Method: Online
Population Sampled: Adults (18+) eligible to vote in the United States
Sample size: 1,500
Weighting: Data were weighted to the profile of adults (18+) in the United States. Data were weighted by age, gender, region, education level, and 2016 Presidential Election Vote. Targets for each weighting were derived from the official estimates of the United States Census and the results of the 2016 Presidential Election.


Gallup AV D+6.6
D 32.5 R 25.9  I 46.6
Results are based on telephone interviews conducted May 28-June 4, 2020 with a random sample of –1,034— adults, ages 18+, living in all 50 U.S. states and the District of Columbia. For results based on this sample of national adults, the margin of sampling error is ±4 percentage points at the 95% confidence level.



HarrisX/The Hill RV D+5
The Hill-HarrisX poll was conducted online among 1,912 registered voters between June 8 and 10. It has a margin of error of plus or minus 2.24 percentage points.
Republican
32%
Democrat
37%
Independent
29%


Meeting Streets Insights RV No DRI This data comes from a Meeting Street Insights online survey of 1,000 registered voters
________________________________________________________





2 comments:

David Herr said...

Hey, great post. I would also add that if one digs into the "I" component, it is actually tilted left, at least it is for those polls that also break down 2016 vote (in other words, I have a higher than normal Clinton 2016 vote level, tilting the entire poll to D even if headline D+ number is within reason). Some polls also have elevated numbers of people who didn't vote in 2016, beyond what one would expect from older voters dying and younger people aging into voting eligibility.

M. Joseph Sheppard said...

That's correct as I pointed out there are a number of factors that can add to a D bias.Look at the PPP methodology for example. It's a Dem biased outfit but hasD+2 report basis but their chart shows Trump Approval landlines 49% "text approval 37%" which arrives at a 43% approval overall which is just silly