Saturday 31 October 2020

That Other Election (Inevitable Post--USA Presidential)

It feels as though, ever since the election of Donald J Trump to the American Presidency, I had no choice but to make a prediction on the 2020 race. So let's get it done now:

Given the level of media coverage this election has received even outside of the US, I will only cursorily describe the electoral college process: votes are held in each of the 50 states and in DC. Each of these states have a number of seats to fill in the electoral college, indirectly linked to their relative populations. For most states, this is on a winner-takes-all basis, so whoever leads in California gets 55 seats electoral college votes (ECVs). There are 538 ECVs and a president needs a majority of these--270. So that's the number we're looking for.

Round 1 - Dividing the Most Divisive

It is not exactly breaking news to suggest political positions have become more polarised in the last four years. The table below compares Trump's vote in 2016 (as a percentage of votes for Trump or Clinton) with the latest polling of likely voters (at 1 November 2020) in each state (again, as a percentage of votes for Trump or Biden).

STATE ECV 2016 LATEST (LV) SHIFT
Alabama 9 64.4% 61.2% -3.1% In Play
Alaska 3 58.4% 54.1% -4.3% In Play
Arizona 11 51.9% 47.9% -4.0% In Play
Arkansas 6 64.3% 61.2% -3.1% In Play
California 55 33.9% 36.7% 2.9% In Play
Colorado 9 47.3% 43.4% -3.9% Biden
Connecticut 7 42.9% 38.4% -4.5% Biden
Delaware 3 44.0% 39.4% -4.6% Biden
District of Columbia 3 4.3% 6.1% 1.8% In Play
Florida 29 50.6% 51.0% 0.4% Trump
Georgia 16 52.7% 49.0% -3.7% In Play
Hawaii 4 32.6% 33.0% 0.4% In Play
Idaho 4 68.3% 61.2% -7.1% In Play
Illinois 20 41.0% 40.8% -0.2% Biden
Indiana 11 60.1% 55.1% -5.0% In Play
Iowa 6 55.1% 50.5% -4.6% In Play
Kansas 6 61.1% 55.1% -6.0% In Play
Kentucky 8 65.7% 59.6% -6.1% In Play
Louisiana 8 60.2% 63.3% 3.1% Trump
Maine 2 48.4% 43.9% -4.5% Biden
Maryland 10 36.0% 32.0% -4.0% Biden
Massachusetts 11 35.3% 29.9% -5.5% Biden
Michigan 16 50.1% 43.6% -6.5% In Play
Minnesota 10 49.2% 30.4% -18.8% Biden
Mississippi 6 59.1% 62.6% 3.5% Trump
Missouri 10 59.8% 56.1% -3.7% In Play
Montana 3 61.1% 53.1% -8.0% In Play
Nebraska 2 63.5% 57.1% -6.4% In Play
Nevada 6 48.7% 49.0% 0.3% In Play
New Hampshire 4 49.8% 47.5% -2.3% Biden
New Jersey 14 42.7% 38.8% -3.9% Biden
New Mexico 5 45.3% 43.9% -1.5% Biden
New York 29 38.2% 35.7% -2.5% Biden
North Carolina 15 51.9% 46.9% -5.0% In Play
North Dakota 3 69.8% 61.2% -8.6% In Play
Ohio 18 54.3% 51.0% -3.2% In Play
Oklahoma 7 69.3% 63.6% -5.7% In Play
Oregon 7 43.8% 39.4% -4.5% Biden
Pennsylvania 20 50.4% 46.3% -4.1% In Play
Rhode Island 4 41.7% 35.7% -6.0% Biden
South Carolina 9 57.5% 55.1% -2.4% In Play
South Dakota 3 66.0% 62.2% -3.7% In Play
Tennessee 11 63.6% 56.6% -7.1% In Play
Texas 38 54.7% 52.0% -2.7% In Play
Utah 6 62.4% 56.1% -6.3% In Play
Vermont 3 34.8% 28.6% -6.2% Biden
Virginia 13 47.2% 42.9% -4.3% Biden
Washington 12 41.2% 36.7% -4.5% Biden
West Virginia 5 72.2% 67.3% -4.8% In Play
Wisconsin 10 50.4% 48.4% -2.0% In Play
Wyoming 3 75.7% 66.3% -9.4% In Play

N.B. Maine and Nebraska are only given 2 ECVs, as these states award two based on state-wide popular votes, with 2 or 3 others respectively awarded based on individual voting districts considered further below.

Shift is the difference between Trump's vote in 2016 and polls in 2020. In seats that he won in 2016 where polling indicates a strengthening of his vote, we can make an early guess that Trump will hold on. These states include Florida which is a key battleground worth 29 ECV or over 10% of a victory. Similarly, in states where Clinton won in 2016 and Biden is polling even better, an early prediction for the Democrats is possible. The rest are still in play.

Round 2 - Deciding the Most Decided

Of course, in some states it's entirely unrealistic to expect a shift of any plausible magnitude to make a difference. To take the most extreme example, Trump's gain in D.C. is the median of all his gains--1.8%; this has reduced Biden's Two-Party percentage to a mere 93.9%.

Polling has a margin of error that can vary a lot depending on sample size, methodology and other factors, but 3 percentage points is a common rule of thumb. Fore safety, I'm allowing a 5% error; seats where Trump holds more than 55% are coloured red and where he has less than 45% go blue in the table below.
 
STATE ECV LATEST (LV)
Alabama 9 61.2% Trump
Alaska 3 54.1% In Play
Arizona 11 47.9% In Play
Arkansas 6 61.2% Trump
California 55 36.7% Biden
Colorado 9 43.4% Biden
Connecticut 7 38.4% Biden
Delaware 3 39.4% Biden
District of Columbia 3 6.1% Biden
Florida 29 51.0% In Play
Georgia 16 49.0% In Play
Hawaii 4 33.0% Biden
Idaho 4 61.2% Trump
Illinois 20 40.8% Biden
Indiana 11 55.1% Trump
Iowa 6 50.5% In Play
Kansas 6 55.1% Trump
Kentucky 8 59.6% Trump
Louisiana 8 63.3% Trump
Maine 2 43.9% Biden
Maryland 10 32.0% Biden
Massachusetts 11 29.9% Biden
Michigan 16 43.6% Biden
Minnesota 10 30.4% Biden
Mississippi 6 62.6% Trump
Missouri 10 56.1% Trump
Montana 3 53.1% In Play
Nebraska 2 57.1% Trump
Nevada 6 49.0% In Play
New Hampshire 4 47.5% In Play
New Jersey 14 38.8% Biden
New Mexico 5 43.9% Biden
New York 29 35.7% Biden
North Carolina 15 46.9% In Play
North Dakota 3 61.2% Trump
Ohio 18 51.0% In Play
Oklahoma 7 63.6% Trump
Oregon 7 39.4% Biden
Pennsylvania 20 46.3% In Play
Rhode Island 4 35.7% Biden
South Carolina 9 55.1% Trump
South Dakota 3 62.2% Trump
Tennessee 11 56.6% Trump
Texas 38 52.0% In Play
Utah 6 56.1% Trump
Vermont 3 28.6% Biden
Virginia 13 42.9% Biden
Washington 12 36.7% Biden
West Virginia 5 67.3% Trump
Wisconsin 10 48.4% In Play
Wyoming 3 66.3% Trump
 
These results are consistent with Round 1, but lock down a lot more predictions (and put Trump's Florida and Biden's New Hampshire back in play.

Round 3 - Determining the Most Determined

Those who can remember 2016 will recall the follies of following polls too closely. Not just the US presidential election, but Brexit and a few other less extreme surprises shook a lot of faith in polls. Certainly, leaning too heavily on a single poll is a risk. Fortunately, 538 lets you view the polling history of each state over a number of months to reduce reliance on outliers.
 
These graphs are corrected for reliability, national trends between state polls, time since last poll etc. as explained here. This is an important point, as it introduces some biases as well as removing them, and can result in some oddities such as Biden never falling behind in Pennsylvania despite Trump having won in an AtlasIntel poll just yesterday (also Trafalgar Group on Oct 29, Insider Advantage Oct 25 and Civiqs Oct 21).

Nevertheless there are a number of states where polling, once adjusted, has without fail indicated one candidate or another would carry the state.

STATE ECV POLLING
Alabama 9 Trump
Alaska 3 Trump
Arizona 11 Mixed
Arkansas 6 Trump
California 55 Biden
Colorado 9 Biden
Connecticut 7 Biden
Delaware 3 Biden
District of Columbia 3 Biden
Florida 29 Mixed
Georgia 16 Mixed
Hawaii 4 Biden
Idaho 4 Trump
Illinois 20 Biden
Indiana 11 Trump
Iowa 6 Mixed
Kansas 6 Trump
Kentucky 8 Trump
Louisiana 8 Trump
Maine 2 Biden
Maryland 10 Biden
Massachusetts 11 Biden
Michigan 16 Biden
Minnesota 10 Biden
Mississippi 6 Trump
Missouri 10 Trump
Montana 3 Trump
Nebraska 2 Trump
Nevada 6 Biden
New Hampshire 4 Biden
New Jersey 14 Biden
New Mexico 5 Biden
New York 29 Biden
North Carolina 15 Biden
North Dakota 3 Trump
Ohio 18 Mixed
Oklahoma 7 Trump
Oregon 7 Biden
Pennsylvania 20 Biden
Rhode Island 4 Biden
South Carolina 9 Trump
South Dakota 3 Trump
Tennessee 11 Trump
Texas 38 Mixed
Utah 6 Trump
Vermont 3 Biden
Virginia 13 Biden
Washington 12 Biden
West Virginia 5 Trump
Wisconsin 10 Mixed
Wyoming 3 Trump

Interval - Consolidated Outputs

These three rounds of consideration align nicely to give us a consolidated prediction for DC and 44 states:
 
STATEECVROUND 1ROUND 2ROUND 3PREDICTION
Alabama9In PlayTrumpTrumpTrump
Alaska3In PlayIn PlayTrumpTrump
Arizona11In PlayIn PlayIn PlayIn Play
Arkansas6In PlayTrumpTrumpTrump
California55In PlayBidenBidenBiden
Colorado9BidenBidenBidenBiden
Connecticut7BidenBidenBidenBiden
Delaware3BidenBidenBidenBiden
District of Columbia3In PlayBidenBidenBiden
Florida29TrumpIn PlayIn PlayTrump
Georgia16In PlayIn PlayIn PlayIn Play
Hawaii4In PlayBidenBidenBiden
Idaho4In PlayTrumpTrumpTrump
Illinois20BidenBidenBidenBiden
Indiana11In PlayTrumpTrumpTrump
Iowa6In PlayIn PlayIn PlayIn Play
Kansas6In PlayTrumpTrumpTrump
Kentucky8In PlayTrumpTrumpTrump
Louisiana8TrumpTrumpTrumpTrump
Maine2BidenBidenBidenBiden
Maryland10BidenBidenBidenBiden
Massachusetts11BidenBidenBidenBiden
Michigan16In PlayBidenBidenBiden
Minnesota10BidenBidenBidenBiden
Mississippi6TrumpTrumpTrumpTrump
Missouri10In PlayTrumpTrumpTrump
Montana3In PlayIn PlayTrumpTrump
Nebraska2In PlayTrumpTrumpTrump
Nevada6In PlayIn PlayBidenBiden
New Hampshire4BidenIn PlayBidenBiden
New Jersey14BidenBidenBidenBiden
New Mexico5BidenBidenBidenBiden
New York29BidenBidenBidenBiden
North Carolina15In PlayIn PlayBidenBiden
North Dakota3In PlayTrumpTrumpTrump
Ohio18In PlayIn PlayIn PlayIn Play
Oklahoma7In PlayTrumpTrumpTrump
Oregon7BidenBidenBidenBiden
Pennsylvania20In PlayIn PlayBidenBiden
Rhode Island4BidenBidenBidenBiden
South Carolina9In PlayTrumpTrumpTrump
South Dakota3In PlayTrumpTrumpTrump
Tennessee11In PlayTrumpTrumpTrump
Texas38In PlayIn PlayIn PlayIn Play
Utah6In PlayTrumpTrumpTrump
Vermont3BidenBidenBidenBiden
Virginia13BidenBidenBidenBiden
Washington12BidenBidenBidenBiden
West Virginia5In PlayTrumpTrumpTrump
Wisconsin10In PlayIn PlayIn PlayIn Play
Wyoming3In PlayTrumpTrumpTrump
 
On these numbers alone Biden would hold 282 ECVs, already a majority and far ahead for Trump's 152. Six states required further consideration, as well as the individual districts in Maine and Nebraska that send more local electors to the electoral college. It is reassuring to note, also, that the arbitrary 5% threshold in round 2 did not define any states that weren't confirmed in another round, making the choice of 5% less critical.

The Outliers

Arizona

Arizona was almost locked in for Biden in Round 3; although Trump has led in this state, that was last seen in early March.
 
Then again, Arizona has only supported two Democratic presidents since the end of World War Two: Truman in 1948 (when the Democrats were still resolving whether they would be the left or right wing party in the USA) and Clinton in 1996 (who won with a landslide 379 ECVs).
 
It is true that an influx of younger people and particularly Latinos and Latinas is shifting the rusted-on red nature of the state, and the push against Trump has to be the best shot the Dems have had in a while, but I'm not convinced the time has come to colour Arizona blue.
 
Arizona Trump

Georgia

Georgia is a very tight race, with Biden and Trump vying for the lead in the polls. At present Biden is up, but like Arizona, the state has a long red streak back decades. The Clinton landslide took the state, as did Georgia's governor Carter twice, but these are obvious outliers as was support for third-place contender George Wallace in 1968.
 
Again, anti-Trump sentiment is a powerful motivator but probably not enough.
 
GeorgiaTrump

Iowa

Iowa is a more contentious state, backing Republicans in 8 of the last 15 presidential elections. Polling is tight but both currently and historically Trump has been looking good here. Could be a surprise Biden win, but the safe bet is against that.
 
IowaTrump

Maine

Main at-large (2 ECVs) has been called for Biden in all three rounds. There are two other votes to be won here, corresponding to Maine's two congressional districts, Maine 1st and Maine 2nd (or as I call them, Main Maine and Maine's Remains).
 
Maine 1st has Biden with a 19 point lead in the most recent polls and the 538 chart shows this has not dropped below 9 points in the last four months of polling.
 
Maine 1st
Biden
 
Maine 2st is much closet with Biden still ahead but only by around 3 points, and a few periods where Trump was leading. This is probably the closest race on my sheets, but at a guess...
 
Maine 2nd
Biden

Nebraska

Like Maine, Nebraska has congressional districts in play: Nebraska 1st, Nebraska 2nd and Nebraska 3rd.

Nebraska 1st has a single poll from August with Trump up 2 points. This isn't a lot to go on, but since Nebraska gave congressional districts individual votes Nebraska 1st has always voted Republican and voted for Trump 56:36 in 2016.
 
Nebraska 1st
Trump

Nebraska 2nd has Biden up 3 points yesterday and by 11 points earlier that month. This seat is the only one to have voted Democrat in Nebraska (for Obama in '08). The polling history suggests that history may repeat itself this year.
 

Nebraska 2nd
Biden
 
 
Nebraska 3rd is historically the most Republican of the three districts. There's no reliable polling here but then, why would there be?

Nebraska 3rd
Trump

Ohio

Ohio is close and Trump has the narrow lead. But the state has also backed the winning president every year sine 1964 (and only four wrong in the century before that). If any state is going to defy the polls, and surely one of them will, it is probably Ohio. For a bit of adventure I'm going to take the minority view here and suggest:

Ohio
Biden

Texas

It's amazing that Texas is in play, and there is some suggestion this may be the new normal as Hispanic communities become a larger proportion of the population. This is a big concern for Republicans, as their biggest safe seat (38 ECVs) is starting to look purple. Still, I think the thought of losing Texas has scared Republicans enough to turn out and vote, and the Trump campaign has had to throw a lot of money into Texas for their own peace of mind; anyone marking Texas blue today is, in my opinion, kidding themselves.

Texas
Trump

Wisconsin

I have 36 October polls from Wisconsin and only one has Trump ahead (and that, on Oct 2, by one point). Eight have Biden up by double digits. What was a close race in April has widened significantly.
 


Wisconsin
Biden
 

Final Prediction 

All combined, that's suggesting:

STATE ECV PREDICTION
Alabama 9 Trump
Alaska 3 Trump
Arizona 11 Trump
Arkansas 6 Trump
California 55 Biden
Colorado 9 Biden
Connecticut 7 Biden
Delaware 3 Biden
District of Columbia 3 Biden
Florida 29 Trump
Georgia 16 Trump
Hawaii 4 Biden
Idaho 4 Trump
Illinois 20 Biden
Indiana 11 Trump
Iowa 6 Trump
Kansas 6 Trump
Kentucky 8 Trump
Louisiana 8 Trump
Maine (at-large) 2 Biden
Maine 1st 1 Biden
Maine 2nd 1 Biden
Maryland 10 Biden
Massachusetts 11 Biden
Michigan 16 Biden
Minnesota 10 Biden
Mississippi 6 Trump
Missouri 10 Trump
Montana 3 Trump
Nebraska (at-large) 2 Trump
Nebraska 1st 1 Trump
Nebraska 2nd 1 Biden
Nebraska 3rd 1 Trump
Nevada 6 Biden
New Hampshire 4 Biden
New Jersey 14 Biden
New Mexico 5 Biden
New York 29 Biden
North Carolina 15 Biden
North Dakota 3 Trump
Ohio 18 Biden
Oklahoma 7 Trump
Oregon 7 Biden
Pennsylvania 20 Biden
Rhode Island 4 Biden
South Carolina 9 Trump
South Dakota 3 Trump
Tennessee 11 Trump
Texas 38 Trump
Utah 6 Trump
Vermont 3 Biden
Virginia 13 Biden
Washington 12 Biden
West Virginia 5 Trump
Wisconsin 10 Biden
Wyoming 3 Trump

That comes to 225 Electoral College votes for Trump and 313 for Biden. At that rate, Biden would be the next President of the United States.