US 2020 Presidential Election Poll Meta-Tracker - Sept 17

(For my latest meta-poll-tracker update, see here)

The US presidential election is of great concern to the entire world. The outcome of the election will have an effect on citizens of numerous countries on matters such as climate change, pandemic response and trade.

Also, with democracy and evidence-based non-authoritarian governance under attack worldwide, the world looks to see evidence that the democratic process can in fact wrestle a country back from a populist authoritarian-leaning leader.

There are numerous polls, and these are tracked by a variety of poll-trackers. It is interesting to see the differences among these trackers. The table below shows those differences and mades its own meta-prediction.

The meta-prediction based on September 7 data is that Joe Biden will win with 333 electoral votes, but could get between 235 and 405.

Biden has to defend his probable lead primarily in 6 states (shown by the green bar below) and also try to increase his margin, with the most likely targets marked with yellow.

Donald Trump has to defend his probable lead in the 5 states highlighted with the yellow bar below, and overcome Biden's probable lead in the 4 states (Arizona, Pennsylvania, Florida and North Carolina) shown with the bold numbers in the green area (or some other combination of states in which Biden is leading, that together have 63 EVs.

The US system isn't based on popular national vote, so Biden could win that with an even higher margin than Hillary Clinton did in 2016 and still lose. Candidates have to focus on obtaining electoral votes that are allocated to each state (one for each senator and representative combined). Other than the two states marked with purple below, the EVs allocated to the state all go to the party that gets the most votes in that state.

The table below is organized as follows.

• The EVs column shows the number of electoral votes each state has.
• The '% prob' column near the right is used for sorting and indicates the precent chance that each party will receive all the electoral votes in that state, except that Democrat wins are shown with negative numbers.
• The 'Cum' column shows the number of electoral votes won, adding the states one by one (Democrats from the top, Republicans from the bottom).
• The 'Certain D' and 'Certain R' states are states (including DC) where all the poll trackers give 99% probability of winning to either party. I have not listed them. That leaves 36 states where at least one poll tracker is not completely certain about either the Democrats or the republicans winning.

The probability calculation is based on the combined weightings of the different poll trackers.

• Two calculations are done. One to compute a minimum number of EVs for each party (second-to-right column, and one for a likely number of EVs (rightmost column). These columns are fractional numbers of EVs because they are probabilistic. Democrat EV counts are shown as negative numbers.
• Each tracker rates some states as near certain wins for a party those are given 95% chance for the minimum EV computation and 100% chance for the likely calculation.
• Each tracker rates another set of states as very likely (VL). Those are given 60% for the minimum and 90% for the likely.
• Each tracker rates a further set of states as Likely (L). These are given 30% and 60% chances, respectively.

• Finally some trackers, but not all, rate some states as toss-ups. These are weighted zero in the calculations.

The poll trackers considered are the following:

• The Economist 'Forecasting the US elections' page: https://projects.economist.com/us-2020-forecast/president This is calculated every day based on state and national polls, and also uses economic indicators. It predicts Biten will get 334 EVs and has an 84% chance of winning currently (although a 96% chance of getting the most votes).
• ElectoralVote.com. https://electoral-vote.com This predicts Biden will win with 350 EVs and also provides news about the election. It seems reasonably neutral, but shows Biden leading in more places than the other trackers.
• The Princeton University Election Consortium https://election.princeton.edu/electoral-college-map/ This uses some slightly different mathematical methods than other trackers, focusing on proabilities rather than simply margins in polls. In other words, greater confidence is given where polls show consistent results, even if the margins might be smaller. It currently predicts Biden will get 366 EVs. 
• Éric Grenier tracker at the CBC in Canada https://newsinteractives.cbc.ca/elections/poll-tracker/presidential/ He currently suggests Biden will get at least 319 EVs, but with 75 as 'toss ups'. 
• Real Clear Politics. https://www.realclearpolitics.com/epolls/2020/president/2020_elections_electoral_college_map.html This tracker differ from the others in that it is far less confident in outcomes. It has 211 EVs and 14 states as toss-ups. It shows Biden as getting at least 212 EVs, but won't go any further than that. I have weighted this tracker at 80% of the weight of the others, due to its lower willingness to make predictions, but it still contributes almost a seventh of the overall weighting.
• FiveThirtyEight. https://projects.fivethirtyeight.com/2020-election-forecast/ This gives Biden a 72% chance of winning (lower than others), and predicts he will get 333 EVs.
• The 2016 results. Although these are old, they are included at 20% of the weighting of the actual trackers to help account for temporal uncertainty (the possibility of people going back to thinking the way they did then, since Trump is well-known for being able to get people to think his way). This only accounts for about 3% of the total.

I will update this data at least a couple of times prior to US Election Day in two months.