Read on Twitter
#2020Election Thread - Battlegrounds Vote - State-by-State and Electoral College.
Background: My 2016 Forecast:
Background: My 2016 Forecast:
My 2016 State-by-State vote and Electoral College Projections relied on Murrey Math and Probable Ranges.
My thought was, rather than landing on a specific number it would be better to focus on probability and a number range. For example 50-53.125%.
My thought was, rather than landing on a specific number it would be better to focus on probability and a number range. For example 50-53.125%.
To forecast 2016 I used the Real Clear Politics (RCP) Average.
The 2016 polls were tilted too heavily in the direction of Hillary Clinton. Polls were off by about 6.25% meaning if a poll said a state race was 50-50 Clinton-Trump I would adjust it 50-3.125% = 46.857% Hillary &
The 2016 polls were tilted too heavily in the direction of Hillary Clinton. Polls were off by about 6.25% meaning if a poll said a state race was 50-50 Clinton-Trump I would adjust it 50-3.125% = 46.857% Hillary &
50+3.125% = 53.125% for Trump. Those numbers would end up being converted to a range.
Hillary at 46.875 would become 46.875 - 50%
Trump at 53.125% would become 53.125 = 56.25%.
There were also some adjustments right before to account for the herding effect.
Hillary at 46.875 would become 46.875 - 50%
Trump at 53.125% would become 53.125 = 56.25%.
There were also some adjustments right before to account for the herding effect.
The herding effect of the tendency of the crowd to herd in the direction of the winner on election day.
This break-down on herding ends up something like -
62.5% of undecideds break for whoever is leading in state polls on Election Day - N(umber) Days.
37.5% of undecideds...
This break-down on herding ends up something like -
62.5% of undecideds break for whoever is leading in state polls on Election Day - N(umber) Days.
37.5% of undecideds...
...break for the loser on Election Day - N Days.
I was able to make my 2016 forecast because I had a fairly plentiful number of polls to look at and had a fair degree of confidence that, as a whole, polls had a sufficient degree of accuracy.
I was able to make my 2016 forecast because I had a fairly plentiful number of polls to look at and had a fair degree of confidence that, as a whole, polls had a sufficient degree of accuracy.
2020 has been more difficult to look at public polling. The polls appear not only more skewed but also less plentiful.
This has made it difficult, if not impossible to put together a forecast using the same methods as 2016.
So, for 2020 I constructed a new model.
This has made it difficult, if not impossible to put together a forecast using the same methods as 2016.
So, for 2020 I constructed a new model.
My 2020 model is based on the following:
1. 6.25% poll bias (just like 2016)
2. Cycles
3. Historical estimations to calculate 3rd Party Vote
4. Herding Effect
The 6.25% poll bias is as described earlier. To correct the bias you have to subtract 3.125% from Biden polls and...
1. 6.25% poll bias (just like 2016)
2. Cycles
3. Historical estimations to calculate 3rd Party Vote
4. Herding Effect
The 6.25% poll bias is as described earlier. To correct the bias you have to subtract 3.125% from Biden polls and...
...add 3.125% to Trump polls.
Cycles/Historical Data looked at 52 vs 56, 68 vs 72, and 80 vs 84 elections.
Estimations of 3rd Party Vote looked at 68-72, 92-96, 2016-Estimate of 2020.
Herding effect was as stated earlier - 62.5% of undecided vote goes to state poll...
Cycles/Historical Data looked at 52 vs 56, 68 vs 72, and 80 vs 84 elections.
Estimations of 3rd Party Vote looked at 68-72, 92-96, 2016-Estimate of 2020.
Herding effect was as stated earlier - 62.5% of undecided vote goes to state poll...
...leader going into the final stretch. 37.5% of undecided vote goes to the loser going into the final stretch.
With that, we can begin to look at some initial projections.
These will be a mix of landing on a number, e.g. 52.3% due to having higher confidence in estimates.
With that, we can begin to look at some initial projections.
These will be a mix of landing on a number, e.g. 52.3% due to having higher confidence in estimates.
Other projections will be in ranges where drilling down on numbers was not as necessary or time did not permit.
Finally, there will be an Electoral College Projection.
Let's begin with some battleground states...
Finally, there will be an Electoral College Projection.
Let's begin with some battleground states...
Current Arizona Trump vs Biden vote projection
Trump - 52.3%
Biden - 45.3%
Trump - 52.3%
Biden - 45.3%
Current Florida Trump vs Biden vote projection
Trump - 52.1%
Biden - 46.3%
Trump - 52.1%
Biden - 46.3%
Current Iowa Trump vs Biden vote projection
Trump - 52.3%
Biden - 44.3%
Trump - 52.3%
Biden - 44.3%
