(C) Daily Kos
This story was originally published by Daily Kos and is unaltered.
 . . . . . . . . . .



Latest Pundit Predictions: Are These Numbers Good or Bad? (with poll). [1]

['This Content Is Not Subject To Review Daily Kos Staff Prior To Publication.']

Date: 2024-09-15  

The purpose of this post is to share my confusion, and some concern, about the current state of the presidential race in regard to pundits and polls, and to get your thoughts on this.

I like to follow several election forecasters, such as fivethirtyeight.com. The numerical predictions they give are based largely on polls. I feel some optimism whenever their predictions look “good” or when their latest numbers look better than their previous numbers. Say, for example (and these numbers are just hypothetical) that one week, the prediction is that Harris will win Michigan by 1.5%, and that in the next week the prediction is that she will win Michigan by 2.5%. I am happy to see this improvement. But I am still confused, and worried, because I don’t know how skewed, and in what direction, the +2.5% is.

Here are the latest predictions, by state, from two respected forecasters, fivethirty.com and racetothewh.com. The seven swing states are listed first, and are followed by several other states where the outcome is at least slightly in question. The numbers represent the percentage amounts by which they predict Harris will win (or lose).

State 538.com racetothewh WI +2.8 +1.83 MI +1.9 +1.79 NV +0.9 +0.83 PA +0.9 +0.77 GA 0 -0.77 NC -0.1 -0.6 AZ -0.1 -0.31 NE-2 +5.9 +5.4 NH +6.6 +6.85 MN +6.9 +8.21 ME +9.2 +9.22 VA +9 +7.79 NM +9.3 +9.19 Alaska -9.7 -8.24 IA -8.8 -6.6 OH -8.1 -9.39 FL -3.8 -4.99 TX -5.4 -6.14 ME-2 -6.4 -4.3 According to the table above, Harris is predicted to win Pennsylvania, Michigan, Wisconsin, and Nevada, which would give her 276 electoral votes. Now, when I look back at what happened in 2020, the numbers above are worrisome. In the table below are the final predictions for the 2020 presidential race, from FiveThirtyEight.com and from RacetotheWH, followed by the actual outcome of the election. PRESIDENT 2020

State fivethirtyeight racetowh Actual Outcome WI +8.3 +8 +0.6 MI +8 +8 +2.8 NV +6.1 +6.1 +2.4 PA +4.7 +5.6 +1.2 GA +0.9 +0.8 +0.3 NC +1.7 +1.9 -1.3 AZ +2.6 +2.7 +0.4 NE-2 +3.2 +3.8 +6.8 NH +10.6 +10.3 +7.3 MN +9.1 +8.8 +7.1 ME +11.8 +12.2 +9.1 VA +12.5 +10.9 +10.1 NM +13 +13.5 +10.8 AK -8.4 -6.2 -10 IA -1.5 -1.3 -8.2 OH -0.6 +0.1 -8.1 FL +2.5 +3.2 -3.3 TX -1.5 -1.5 -5.6 ME-2 +0.6 +2 -7.9

From the above table, we see that both forecasters greatly overestimated Biden’s performance in the election. In fact, the only contest that they did not overestimate for Biden was in Nebraska’s 2nd district. In particular, if we consider only the 7 battleground states, FiveThirtyEight.com’s average error was +3.7 in favor of Biden, and RTTWH’s average error was +3.81 in favor Biden. If we take the entire list of states and districts in the above table, then the average errors were +3.63 and +3.93, respectively.

What worries me is this: if the poll were to have about the same degree of bias in 2024 as they had in 2020, then, based on the current polls, the outcome for 2024 would be a disaster, as we would lose every swing state.

HOWEVER, maybe the pollsters and/or the prognosticators will not be off as much, or in the same direction as in 2020. In fact, in 2022, looking at the senate races, the bias of their predictions turned out be quite the opposite of what happened in 2020.

Here are the 2022 Senate race final predictions (for the somewhat competitive races), and the outcomes. The numbers indicate the percentage advantage for the Democrat.

SENATE 2022

State Fivethirtyeight Racetothewh Actual Outcome AZ +2.2 +1.9 +4.9 COL +8 +8.9 +14.6 GA -1.2 +0.5 +2.8 FL -9.8 -9 -16.4 NC -4.8 -4.9 -3.2 NH +3.5 +4.6 +9.1 NV -0.2 -0.3 +0.8 OH -6.4 -5.4 -6.1 PA -1 +0.2 +4.9 WA +8.6 +6.6 14.6 WI -4.8 -6 -1

From the above table, we see that in almost of the races, both forecasters underestimated the performance of the democrats. If we consider only the five battleground states that are at stake in this year’s presidential election, the two forecasters were off by an average of 3.04 and 2.56 points, respectively, in favor of the Republican. When we add in all of the other states in the table, then the forecasters’ average errors were 2.46 and 2.08 points, respectively, in favor of the Republican.

So, in sharp contrast to the bias in the predictions of the 2020 election, the 2022 predictions had the oppositive bias. Now, if we use the same degree of bias in 2024 as that of the 2022 Senate races, then, based on the current polls, Harris should win all 7 of the swing states (and no others), giving her 319 electoral votes.

As one final comment here, a natural question would be: even if we assume that this year’s predictions behave more like those of 2022 than those of 2020, is there any validity to comparing Senate race forecast bias with Presidential race bias? In trying to answer, I think it would be worth looking at the 2020 Senate predictions to see how they compare to the 2020 Presidential and 2022 Senate forecasts. So, here is my last table.

SENATE 2020

State 538.com RaceToThewh.com Actual Outcome AK -5.8 -3.1 -12.71 AZ +5.2 +5.5 +2.35 CO +7.7 +9.3 +9.32 GA -0.3 +1.1 +1.22 IA -1.4 +0.3 -6.59 KS -5.8 -3.4 -11.43 ME +2 +4.8 -8.59 MI +7 +8 +1.68 MN +12.3 +9.1 +5.24 MT -3.2 -2 -10.02 MS -9.2 -4.2 -9.97 NC +3.2 +3.7 -1.75 NM +12.9 +10.9 +6.11 SC -5.1 -3.8 -10.27 TX -7.3 -5.8 -9.64

So, the 2020 Senate predictions were off to a similar degree as the 2020 Presidential predictions. Just looking at the battleground states, the 2 forecasters had an average bias in favor of the Democrats of 2.9, and 3.7, respectively. And if we include all 15 states in the table, the biases for the Senate predictions were 4.48 and 5.7, respectively.

In conclusion: In 2020, the outcomes were much more in favor of the Republicans than had been predicted, while in 2022, the outcomes turned out much better for the Democrats than had been predicted. I very much hope that, in 2024, the actual results, when compared to the predictions based on polls, will turn out like those of 2022, and not like those of 2020. If so, I believe we will win!

[END]
---
[1] Url: https://www.dailykos.com/stories/2024/9/15/2270488/-Latest-Pundit-Predictions-Are-These-Numbers-Good-or-Bad-with-poll?pm_campaign=front_page&pm_source=more_community&pm_medium=web

Published and (C) by Daily Kos
Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/dailykos/