Exit Polls and Pre-Election Polls Versus Computer Counts
By Theodore de Macedo Soares
Exit polls and pre-election polls projected the Democratic Party to gain control of the Senate on election day 2020. The computerized vote counts have the Republican Party with a two seat advantage. Like the 2020 and 2016 presidential elections and the 2016 U.S. Senate elections the 2020 US Senate races show almost all discrepancies, between the exit polls and the unverified computer counts, favored the Republican Party candidate. Most discrepancies were large and beyond the margin of error for the exit poll. See tables below.
Edison Research conducted exit polls on the U.S. Senate races in 16 states. In this election, exit polls projected the Democratic Party to gain control of the Senate with 51 seats to 49 for the Republican Party. The pre-election polls projected 52 seats for the Democratic party and 48 seats for the Republicans. As Georgia’s two Senate contests did not result in any candidate achieving 50% of the vote, the runoff elections on January 5, 2021 will determine the final Senate seat apportionment between the two main political parties. The computerized vote counts currently show the Republican Party with 50 seats to 48 for the Democratic Party.
The exit polls conducted by Edison in this coronavirus pandemic year used the same methodology as in previous years. In 2016 absentee and early voting represented about 40% of the votes, this year it will exceed 60%. The fact that for the 2016 Republican Party primaries the vote counts matched the exit polls—with most misses within the margin of error for the polls and randomly distributed for and against candidate Trump—provide evidence that exit polls can be highly accurate.
As previously noted, the wholesale one-way variances between the exit and pre-election polls with the computerized vote counts, all favoring the Republican candidate, are a near statistical impossibility. As the source of the problem is not due to random variations its origin must be systemic. Either exit polls and pre-election polls are improperly, election after election, conducted or the computerized vote counts corrupt.
Observations on Some U.S. Senate Races
Maine – The most noteworthy aspect in the Senate race between Sara Gideon (Dem) and longtime incumbent Susan Collins (Rep) occurred with the pre-election polls. Not a single poll, of the 14 polls listed by RealClearPolitics since February 2020, gave Collins a lead. Not a single poll of the 22 polls compiled by FiveThirtyEight since July 6, 2020 gave Collins a lead. Gideon led in every poll—RealClearPolitics’ polls since the beginning of October averaged a lead of 5.7%, 538’s averaged a lead of 4.7% (listed in tables below). The 12.6% discrepancy between 538’s compiled polls and the computer vote count was the highest of all the US Senate races listed in the tables below.
North Carolina – In 65 polls compiled by FiveThirtyEight since September 1, 2020 Cal Cunningham, Democrat led in 62. Republican incumbent Thom Tillis only led in one, with two polls tied. Similarly 43 polls listed by RealClearPolitics since September 1 Tillis led in only two with two tied. The unverified computer vote counts, however, declared Republican Thom Tillis the winner.
Georgia – The state had two Senate elections. The discrepancy between the exit poll and the vote count in the Senate race between the incumbent David Perdue (Rep) and Jon Ossof (Dem) favoring Perdue exceeded the margin of error for exit poll. See table 2. The other, a US Senate Special Election to fill a seat temporarily held by Senator Kelly Loeffler (Rep) who was appointed last year. The two exit poll results for the Democratic Party candidates, Raphael Warnock and Matt Lieberman, and the two from the Republican Party, Kelly Loeffler and Doug Collins, are combined in table 2 by their respective party. The discrepancy favoring the Republican Party candidates far exceeded the margin of error for the exit poll. See table 2.
Alabama – Although both the exit and pre-election polls had Jones (Dem) losing to Tuberville (Rep) this state had the distinction of having the highest discrepancy between the exit polls and the computer vote count at 13.2% and the third highest with pre-election polls at 10.0%. Alabama was the only U.S. Senate seat flipping from the Democratic to the Republican Party.
The United States remains one of the few major democracies in the world that continue to allow computerized vote counting—not observable by the public—to determine the results of its elections. Countries such as Germany, Norway, Netherlands, France, Canada, United Kingdom, Ireland, Spain, Portugal, Italy, Denmark, Sweden, Finland and many other countries protect the integrity and trust of their elections with publicly observable hand-counting of paper ballots.
 Exit polls (EP) conducted by Edison Research and published by CNN shortly after the closing of polls for the state and downloaded by TdMS. Edison Research conducted EPs for Senate races in 16 states. Exit poll results are derived from the gender category–the proportion of men and women voting for each candidate. As these first published exit polls are later altered/adjusted to conform with the unverified computer vote counts, the discrepancies shown above are adjusted to near zero in the final EPs.
 New York Times reported vote counts. https://nyti.ms/3mN2ujW
 The margin columns subtract the Republican Party candidates’ vote percentages from the Democratic Party candidates. A Democratic Party win is shown by a positive sign and a Republican Party win by a negative sign.
 Note that the Margin of Error (MOE) is for the differences between the two candidates (at 95% CI). This MOE is about double the usual MOE for each candidate. MoE calculated with multinomial formula discussed in sections 2 and 4 in: Franklin, C. The ‘Margin of Error’ for Differences in Polls. University of Wisconsin, Madison, Wisconsin. October 2002, revised February 2007. Available at: https://abcnews.go.com/images/PollingUnit/MOEFranklin.pdf
 Georgia held a U.S. Senate Special Election to fill a seat temporarily held by Senator Kelly Loeffler (Rep) who was appointed last year. The exit poll results for the four candidates listed in the poll, two from the Democratic Party, Raphael Warnock and Matt Lieberman, and two from the Republican Party, Kelly Loeffler and Doug Collins, are combined in the table above by their respective party. Their computer vote counts have also been combined by party. Since no candidate received the majority of votes, the top two finishers, Warnock and Loeffler, will hold a runoff election on January 5, 2021.  In Georgia’s other Senate race between the incumbent David Perdue (Rep) and Jon Ossof (Dem), neither received over 50% of the vote, they will also participate in a runoff election on January 5, 2021. These two Georgia seats will determine which party controls the Senate.
 The three “margin” columns subtract the results of the exit polls, pre-election polls, and computer vote counts of the Republican Party candidate from the Democratic Party candidate. On the exit poll column, for instance, in Georgia, The Democrat exit poll’s percentage was 46.2% and the Republican’s 53.4% for a difference of -7.2% (46.2% – 53.4%). In Montana, the Democrat received 50.4% and the Republican 49.1% for a difference of 1.3% (50.4% – 49.1%). See first three columns in Table 1 above. Thus, a negative sign shows a lead by the Republican candidate and a positive sign a lead by the Democrat.
Similarly, the Pre-Election Poll column shows the difference (Democrat average poll % – Republican average poll %) between the two candidates. Same for the Computer Vote Counts column (see three columns under Reported Vote Count in Table 1 above).
 This column shows the difference between the exit poll results of the Democrat versus Republican candidate (Dem exit poll % – Rep exit poll %). The exit poll results for the Democratic and Republican candidates are shown in first three columns in Table 1 above.
 This column shows the difference between the pre-election poll results of the Democrat and Republican candidate (Dem average of pre-election polls % – Rep average of pre-election polls %). Pre-election polls used here were compiled by FiveThirtyEight.com (click on “state” drop-down box). The poll results above used the following procedures:
a) Only polls conducted entirely in October through November 2 are included. (Negligible differences if limited to October 15 on)
b) Only polls of “likely voters” were used—almost all the polls. Polls based on “registered voters” or “adults” excluded
c) Polls, such as by Monmouth University, that applied differing assumptions to the same poll to come up with three or so poll results were averaged into one result
d) Two pollsters were excluded. Polls by swayable.com conducted entirely online and tending to bias towards the Democratic Party candidate were excluded. Polls by thetrafalgargroup.org were excluded as opaque and not reliable
 This column shows the difference between the vote count proportion of the Democrat versus Republican candidate (Dem vote count % – Rep vote count %). The vote count proportions for the Democratic and Republican candidates are shown in Table 1 above. Vote proportions based on the current computer vote counts, as of the date below the table, and supplied by the New York Times
 The two discrepancy columns subtract the computer vote count margins from the exit poll and pre-election poll columns. In Alabama, for instance, the exit poll column shows a Republican lead of 7.2% (negative sign on the table because smaller Dem % – greater Rep %). The vote count column shows a Republican lead of 20.4% resulting in a huge 13.1% discrepancy in favor of the Republican candidate.
In Maine, the average of the pre-election polls had the Democrat winning by 4.7% (remember, a positive sign means a Democrat lead) and yet the Vote Count column gave a lead of 8.6% to the Republican candidate so the last column shows a huge discrepancy of 13.3% in favor of the Republican candidate (4.7% – (-8.6%))
December 16, 2020. Added links to the actual exit polls used in this article with an explanation on how the exit poll proportions were derived. Added clarifying notes to Table 1 to make it more understandable.
December 17, 2020. Changed the presentation order of the two tables.
Exit polls downloaded from CNN’s website at poll’s closing: ALABAMA – ARIZONA – COLORADO – GEORGIA 1 – GEORGIA 2 – IOWA – KENTUCKY – MAINE – MICHIGAN – MINNESOTA – MONTANA – NEW HAMPSHIRE – NORTH CAROLINA – OREGON – SOUTH CAROLINA – TEXAS – VIRGINIA. ALL 2020 EXIT POLLS (31MB Zip file)
The exit poll vote proportions were derived from the gender category (all the other categories would have approximately the same result). Democrat’s proportion of the male vote was multiplied with the total male proportion and added to Democrat’s proportion of the female vote multiplied with the total female proportion to arrive at Democrat’s exit poll vote share in the state. The same procedure was applied to arrive at Republican’s proportions of the exit poll vote.
Please credit this article if you use these exit polls or figures derived from them in your publications.