Category Archives: Election Analysis

The Early Vote Predictor

By Jim Ellis

Dec. 3, 2020 — While it became clear that the plethora of polling data published during the 2020 election cycle again proved to be a poor predictor of eventual campaign outcome in most states, another resource was discovered that might be the more reliable prognosticator.

The states releasing early voting numbers before the election – not the actual election results, of course, but the numbers of Democrats, Republicans, and Unaffiliated voters who had returned their ballots – provided the opportunity of charting possible race outcomes. As a predictor, the early voting numbers, largely because they are actual votes and not extrapolations and estimates as found in polling, look to be a more reliable gauge.

Let’s examine the results in the key battleground states and compare them to both the 2020 and 2016 early voting numbers as compiled by the Target Smart organization. Target Smart monitored, categorized, and published the early ballot return numbers throughout the acceptance period in every state that publicly released such data. Unfortunately, for purposes of our exercise, the Georgia 2020 numbers are among the states not currently available.

In Arizona, we see a difference in the 2016 and 2020 early vote numbers that indicated a small shift in the voting patterns. Detecting that Democratic early vote participation had increased several points from four years ago while Republicans were down slightly did prove indicative in relation to the final Arizona result that yielded an official 10,457 vote Biden victory.

In Florida, we see the Democratic early vote numbers dropping slightly. This is a bit surprising in that 2020 featured a record voter turnout. The fact that Republicans gained a bit in the swing was a predictor of President Trump’s stronger performance in the Sunshine State as compared to the result from four years previous.

As we can see from the Michigan numbers, Democrats increased their early voting participation while Republicans saw a decrease. Unaffiliated voters substantially increased. Considering the final result, it is apparent that most of the Unaffiliateds voted Democratic in the presidential contest.

The North Carolina early vote numbers gave us our first clear indication that the pollsters were mis-casting the state’s electorate. The clear indication that Democrats were missing their marks in early voting while Republicans were exceeding their expectations was the first indication that the final vote would produce a different result than the plethora of polls were suggesting.

Pennsylvania featured drastic changes in not only the partisan early vote pattern, but also in volume as early voting increased by more than twelve-fold in comparison to 2016. The stark difference in Democratic versus Republican participation levels did forecast a swing to the former party, though the final totals were not as drastically different as the early voting yields.

Continue reading

Analyzing the Patterns

By Jim Ellis

President Trump | via Flickr

Dec. 2, 2020 — Now that election results are being certified around the country, we can begin to analyze the numbers in an attempt to detect what voting patterns developed throughout the electorate.

In looking at the presidential state-by-state totals from all 50 states and the District of Columbia, we can begin to see that President Trump fell below his previous vote marks not only in places like Arizona, Georgia, and the Great Lakes region, but in several other places, as well. This, despite seeing over 10.5 million more people voting for him in 2020 when comparing his totals from those recorded four years ago.

In a total of 18 states, Trump dropped below his 2016 performance rate, including eight places that he carried in both 2016 and 2020. In all eight, however, his drop-off rate was less than one-half percentage point.

Conversely, in 32 states, he exceeded his 2016 performance mark and surprisingly so in such left of center states as California (+2.7 percent), Hawaii (+4.9 percent), Nevada (+2.2 percent), New Mexico (+3.5 percent), New York (+5.6 percent), and Washington (+2.2 percent). Mind you, he came nowhere near carrying any of these states, with the exception of Nevada, but the president did record slight improvement when compared with his 2016 vote performance.

The state where Trump outperformed his 2016 total by the most is Utah (12.6 percent), but that is largely because there was no strong Independent or significant minor party candidate on the ballot in the 2020 election. Four years ago, Independent Evan McMullen did well in Utah, attracting 21.5 percent of the Beehive State vote thus allowing Trump to carry the state with only a plurality of 45.5 percent. In 2020, his victory percentage improved to 58.1 percent.

One of the key reasons former vice president Joe Biden won the election is because he increased Democratic performance over Hillary Clinton’s 2016 vote totals in every state but New York. This allowed him to tip the balance away from President Trump in the critical states of Arizona, Georgia, Michigan, Wisconsin, and Pennsylvania, those states the latter man carried in 2016 but were lost to Biden in this current election.

Continue reading

Final House Tally: 223-212?

By Jim Ellis

Dec. 1, 2020 — The latest outstanding congressional race numbers suggest that the House may break 223 Democrats to 212 Republicans when some very close elections are finally decided. If this is indeed the final party division among the 435 seats, the GOP will only be six congressional districts away from re-claiming the House majority in the 2022 elections.

Currently, we see Real Clear Politics projecting California Rep. Mike Garcia (R-Santa Clarita) as the winner over state Assemblywoman Christy Smith (D-Newhall) in the state’s 25th CD, which lies in parts of both Los Angeles and Ventura Counties.

With still a small undetermined number of votes to be verified and counted, Garcia’s tight 405-vote margin appears to be holding. Our own rudimentary projections suggest that the freshman Republican congressman will hold by just under 400 votes. It is probable we will see a recount and potential ballot challenges, so the result may be challenged before and after certification. Under California election law, certification for all races must occur by Dec. 11.

We have seen several projections made suggesting former California Rep. David Valadao (R-Hanford) has defeated freshman Rep. T.J. Cox (D-Fresno) in the state’s Central Valley 21st CD. Valadao has a 1,820-vote lead according to the California Target Book organization’s in-depth analysis.

Our more rudimentary estimates suggest that Valadao will win the final count by approximately 1,400 votes based upon the potential number of outstanding ballots in the three counties, Kern, Kings, and Tulare, that are still verifying and counting mail votes. Fresno County has completed its count.

Later today, we expect to see the second certification of Iowa candidate Mariannette Miller-Meeks (R) in the state’s 2nd Congressional District. Under Iowa law, a certification had to be issued one week after the election, in this case Nov. 10, and Miller-Meeks, a state senator from Ottumwa who is in her fourth run for the US House, was originally certified as the winner with a 47-vote margin. The full recount finds former state senator and 2018 lieutenant governor nominee Rita Hart (D) gaining votes but still losing by a total of just six votes districtwide from more than 394,400 ballots cast, a Miller-Meeks winning percentage of 50.0008 percent.

Continue reading

Presidential Polling Report – Part II

By Jim Ellis

Nov. 30, 2020 — Last week, we covered the cumulative polling community’s 2020 accuracy in the Great Lakes region (scroll down below), and today we look at how their predictions fared in President Trump’s five core states of Arizona, Florida, Georgia, North Carolina, and Texas. If the President had been able to carry all five states, and he has now officially failed in Arizona and Georgia, he would have only needed to win one of the battleground state bordering a Great Lake to secure re-election.

In each place, we take the five publicly released polls conducted closest to the election and analyze whether the cumulative and individual survey research firms came close to the final result or missed beyond the polling margin of error.


Arizona

Canvassing is not yet complete in Arizona, but it appears Mr. Biden scored a 49.4 to 49.1% victory, a margin of just 10,457 votes. Carrying Arizona was the first conversion step for Mr. Biden to win the national election.

The final five pollsters were the NBC News/Marist College; CNBC/Change Research; Reuters/Ipsos, Emerson College, and Rasmussen Reports. Here, NBC/Marist came the closest, predicting the two candidates landing in virtually a dead heat. All five, however, were within the polling margin of error, though Rasmussen Reports did call the winner wrong, predicting President Trump would carry the state by four percentage points.

Arizona – Biden 0.3 percent

NBC News/Marist 10/29 – 11/1 717 LV 48 48 Tie
CNBC/Change Research 10/29 – 11/1 409 LV 50 47 Biden +3
Reuters/Ipsos 10/27 – 11/1 610 LV 49 47 Biden +2
Emerson 10/29 – 31 732 LV 48 46 Biden +2
Rasmussen Reports 10/27 – 29 800 LV 45 49 Trump +4

Florida

Cumulative polling predictions again projected a close Democratic win during most of the late election cycle only to see, for the fourth consecutive time in a major Florida statewide race, the Republican candidate reversing the trend and clinching a close win. The latest Florida data was much closer to the mark as three of the final five pollsters correctly forecast a win for President Trump.

The Trafalgar Group proved the closest with their Trump +2 final projection. CNBC/Change Research missed by the most, a six-plus point swing from their Biden +3 prediction to a Trump +3.3 final result.

The final five Sunshine State pollsters were: Fox 35/Insider Advantage; the Trafalgar Group; CNBC/Change Research; Susquehanna Polling & Research; and Rasmussen Reports.

Florida – Trump 3.3 percent

FOX 35/Insider Adv 11/2 400 LV 47 48 Tie Trump +1
Trafalgar Group 10/31 – 11/2 1003 LV 47 49 Tie Trump +2
CNBC/Change Research 10/29 – 11/1 806 LV 51 49 Tie Biden +3
Susquehanna 10/29 – 11/1 400 LV 46 47 Tie Trump +1
Rasmussen Reports 10/29 – 31 800 LV 48 47 Tie Biden +1

Georgia

Just about everyone missed the final count in Georgia, though pollsters throughout the election cycle were clearly forecasting a close Peach State race that certainly ended in such a manner. With the race now certified, Biden won the state, 49.5 to 49.3 percent, a margin of just 12,670 votes.

The only pollster to correctly predict a Biden victory was Public Policy Polling, while the Trafalgar Group, despite being closest in Arizona, was furthest away in Georgia. The five pollsters were: WSB-TV/Landmark Communications; the Trafalgar Group; Insider Advantage; Emerson College; and Public Policy Polling.

Georgia – Biden 0.2 percent

WSB-TV/Landmark 11/1 500 LV 50 46 Trump +4
Trafalgar Group 10/31 – 11/2 1013 LV 50 45 Trump +5
Insider Advantage 11/1 500 LV 48 46 Trump +2
Emerson 10/29 – 31 749 LV 49 48 Trump +1
PPP 10/27 – 28 661 LV 46 48 Biden +2

North Carolina

North Carolina, being one of the quintessential swing states in the country, again produced a close race in 2020 as President Trump claimed the Tar Heel State with just a 1.3 percentage margin, 49.9 – 48.6 percent. Rasmussen Reports was again closest to the mark, as they were in Pennsylvania. Most of the polling throughout the entire election cycle, however, 57 of 85 published polls with eight ties, forecast Biden as holding a North Carolina lead.

The five final pollsters were: Insider Advantage; CNBC/Change Research; Emerson College; Reuters/Ipsos; and Rasmussen Reports. While RR was the closest, CNBC/Change Research was furthest away.

North Carolina – Trump 1.3 percent

Insider Advantage 10/30 – 31 450 LV 48 44 Trump +4
CNBC/Change Research 10/29 – 11/1 473 LV 47 49 Biden +2
Emerson 10/29 – 30 855 LV 47 47 Tie
Reuters/Ipsos 10/27 – 11/1 707 LV 48 49 Biden +1
Rasmussen Reports 10/28 – 29 800 LV 48 47 Trump +1

Texas

Media report after media report continued along the theme that the Lone Star State of Texas was in play for Biden but, in the end, Texas remained red and voted for President Trump in a 5.5 percentage spread, 52.0 – 46.5 percent, which was the closest major statewide race in nearly two decades.

At the end of the race, the pollsters were detecting a clear move toward President Trump largely because the oil and gas industry issues, so important to the Texas economy, became the focal point of attack throughout most of the campaign.

The University of Houston poll called the race almost exactly correct, while two academic institution pollsters, Emerson College and Quinnipiac University, were furthest away. The five pollsters were: Emerson College; University of Massachusetts at Lowell; Siena College/New York Times; Quinnipiac University; and the University of Houston.

Texas – Trump 5.5 percent

Emerson 10/29 – 31 763 LV 49 49 Tie
UMass Lowell 10/20 – 26 873 LV 48 47 Trump +1
Siena College/NYT 10/20 – 25 802 LV 47 43 Trump +4
Quinnipiac 10/16 – 19 1145 LV 47 47 Tie
University of Houston 10/13 – 20 1000 LV 50 45 Trump +5

ME-2

For the second time in two presidential elections, the 2nd Congressional District of Maine bucked the statewide trend and supported President Trump. Biden’s larger margin in the ME-1 CD allowed him to carry the statewide count, which earned him three of Maine’s four electoral votes.

As in the US Senate election in which the polling community missed the final result by the largest margin in the country, so too did they underestimate President Trump’s strength in northern Maine.

All five final pollsters forecast a Joe Biden victory in ME-2, but the end result was a substantial win for President Trump. Each missed well beyond the polling margin of error. The five polling firms were: Change Research; Emerson College; Survey USA; Colby College; and Pan Atlantic Research.

ME-2 – Trump 7.9 percent

Change Research 10/29 – 11/2 475 LV 47 46 Biden +1
Emerson College 10/29 – 31 301 LV 50 47 Biden +3
Survey USA 10/23 – 27 509 LV 48 45 Biden +3
Colby College 10/21 – 25 453 LV 46 42 Biden +4
Pan Atlantic Research 10/2 – 6 300 LV 47 43 Biden +4

Presidential Polling Report – Part I

By Jim Ellis

Note: All the best for a wonderful Thanksgiving holiday. Political Updates will return on Monday, Nov. 30.

Former VP Joe Biden

Nov. 25, 2020 — Now that the states are certifying their election results and the numbers are becoming clearer, we can look at the key battleground regions and assess the cumulative polling community’s accuracy.

Today we look at the key Great Lakes region and the states that turned away from President Trump and landed in former vice president Joe Biden’s camp. Polling had predicted Biden to win all four of the key states, Michigan, Minnesota, Pennsylvania, and Wisconsin, which he did, but they largely missed the mark when it came to margin. In three of the five instances the pollsters predicted substantial wins for Biden, not the close result he ultimately recorded.

In each place, we take the five publicly released polls conducted closest to the election and analyze whether the cumulative and individual survey research firms came close to the final result or missed beyond the polling margin of error.


Michigan

In Michigan, where Biden scored a 50.6 to 47.8 percent victory, the final five pollsters were the Trafalgar Group, Insider Advantage, CNBC/Change Research, Emerson College, and Mitchell Research. Insider Advantage came the closest, predicting a Biden two-point victory. Three pollsters missed beyond the polling margin of error. The Trafalgar Group actually predicted a Trump win.

Michigan – Biden 2.8 percent

Trafalgar Group 10/30 – 31 1033 LV 46 48 Trump +2
Insider Advantage 10/30 – 31 500 LV 49 47 Biden +2
CNBC/Change Research 10/29 – 11/1 383 LV 51 44 Biden +7
Emerson 10/29 – 30 700 LV 52 45 Biden +7
Mitchell Research 10/29 817 LV 52 45 Biden +7

Minnesota

Though Minnesota is the most loyal state to the Democrats in the presidential race, it appeared for a time that the state could become close, just like in 2016 when President Trump came within 1.5 percentage points of capturing the domain. In 2020, however, Minnesota returned to form and awarded Biden a 52.5 to 45.4 percent win. Here, the final five pollsters were Survey USA, the Trafalgar Group, KSTP/Survey USA (two separate polls), and Minnesota Post. All fell within an acceptable accuracy range.

Minnesota – Biden 7.1 percent

SurveyUSA 10/23 – 27 649 LV 47 42 Biden +5
Trafalgar Group 10/24 – 25 165 LV 48 45 Biden +3
KSTP/SurveyUSA 10/16 – 20 625 LV 48 42 Biden +6
MinnPost 10/12 – 15 1021 LV 49 44 Biden +5
KSTP/SurveyUSA 10/1 – 6 929 LV 47 40 Biden +7

Pennsylvania

The Keystone State has been the site of the most strenuous voter fraud lawsuits, which included a Trump legal victory. In the end, however, the victorious lawsuit(s) would not be enough to overturn the projected result, which was a close 50.0 – 48.8% finish. Polling was inconsistent with Rasmussen Reports coming closest to the final result. Three of the five firms actually predicted a Trump victory. The pollsters were: Susquehanna Polling & Research, Rasmussen Reports, the Trafalgar Group, NBC News/Marist College, and Insider Advantage.

Penn – Biden 1.2 percent

Susquehanna 11/1 – 2 499 LV 48 49 Trump +1
Rasmussen Reports 10/31 – 1 800 LV 50 47 Biden +3
Trafalgar Group 10/30 – 31 1062 LV 46 48 Trump +2
NBC News/Marist 10/29 – 11/1 772 LV 51 46 Biden +5
Insider Advantage 10/30 – 31 500 LV 47 49 Trump +2

Wisconsin

The Badger State was the closest of the four regional entities that touch a Great Lake. The final result saw the two candidates coming within just 20,608 votes, or a percentage breakdown of 49.4 – 48.8 percent in favor of the Democratic nominee. Polling from four of the five entities missed badly, all predicting a substantial win for Biden.

Susquehanna was the closest of the group, missing by just over two percentage points. CNBC/ Change Research, Susquehanna Polling & Research, Emerson College, Ipsos/Reuters, and Siena College/New York Times comprised the final group of survey research entities.

Wisconsin – Biden 0.7 percent

CNBC/Change Research 10/29 – 11/1 553 LV 53 45 Biden +8
Susquehanna 10/29 – 31 450 LV 49 46 Biden +3
Emerson 10/29 – 30 751 LV 53 45 Biden +8
Reuters/Ipsos 10/27 – 11/1 696 LV 53 43 Biden +10
NY Times/Siena 10/26 – 30 1253 LV 52 41 Biden +11

NE-2

For the second time in the past four presidential elections, the 2nd Congressional District of Nebraska went against the statewide trend and awarded an electoral vote to the candidate losing the statewide count. In addition to this year, the same pattern occurred in 2008 when then-Sen. Barack Obama carried the district but not the state.

This time the Biden victory was substantial with FM3 Research projecting the correct margin of 11 percentage points. The other pollsters, while forecasting a Biden win, significantly missed the actual winning spread. The five pollsters were: UNLV Business School, Change Research, Emerson College, FM3 Research, and Siena College/New York Times.

NE-2 – Biden 11.4 percent

UNLV Business School 10/30 -11/2 191 LV 50 44 Biden +6
Change Research 10/29 – 11/2 547 LV 50 47 Biden +3
Emerson College 10/29 – 30 806 LV 50 48 Biden +2
FM3 Research 10/1 – 10/4 450 LV 53 42 Biden +11
Siena College/NYT 9/25 – 9/27 420 LV 48 41 Biden +7