F11-F14 Random stuff Flashcards
What could be reasons for polls over- or underestimating Trump?
Overestimating: Overcorrect for past mistakes
Underestimate: Trump mobilize voters that are difficult to capture in a poll (partisan nonresponse). Does the actual turnout underrepresent trump (population effects)
What could be reasons for Trump winning?
The bro-strategy worked
What could be reasons for Harris winnig?
A massive gender gap (as predicted) / strong female vote
Low turnout of non-college
What could be reasons for polls over- or underestimating Harris?
Overestimating: Lower turnout than predicted. Herding?
Underestimating: Recalled 2020 vote adjustments
Why shouldn’t we trust forecasts?
Grimmer et al. (2024) show that we are decades away from assessing probabilistic forecasting provides reliable insights into election outcomes due to limited empirical data (elections).
What could a solution to assessing model performance in other ways than ‘right or wrong’
Morris: Looking a the distance from the result on vote share
What is MSE?
Mean squared error. The average squared difference between observed and predicted value.
MSE is always positive and sensitive to outliers.
What type of election was 2024?
Maybe more balancing than realignment. A realignment would have been clearer in polls
Why was the turnout lower?
No primaries could be reasons for less mobilization.
How was the national swing?
Very uniform towards Trump nation wide expect a few places (GA amongst others).
What could lead to polling error in 2024?
Closet republicans (not voting Harris)
Changing of the latino vote
Adjustment for recalled vote
Less incentive to get ride of Trump (very strong motivation in primaries for Biden).
The bro vote (Elon Musk etc.)
What was the polling error in the last 21 days according to 538?
2024: D+2.7
2020: D+4.2
2016: D+3.2
Which states had big/small polling error in 2024
Iowa: D+8.2
Florida: D+6.6
Georgia: D+1.0
Wisconsin: D+1.6
How was the polling error in swing states?
All overestimated Democrats between +1 and +3
Why is the assumption of correlation in state-level sample error not the best?
Past three elections:
Iowa between Dem+6.2 and Dem+8.2
Georgia between Dem+0.7 and Dem+1.4
Explain anti-incumbency explanation of the eleciton.
All incumbents have lost in 2024. Covid-19 and inflation backlash?
Explain partisan nonresponse bias in 2024 polls
An explanation could be that Trump mobilizes republican voters with the lowest propensity score for voting (young men etc.)
What about the latino vote in 2024?
One county in Texas with 97 pct. latino shifted 75 pct. points Trump from 2016 to 2024.
What was special about turnout in 2024?
People had hadn’t voted since 2018 overwhelmingly voted Trump. Maybe he mobilizes people because of inflation that normally doesn’t vote.
What can we say about polls in general in 2024?
Fewer polls but with higher quality.
Fewer polls due to late change in candidate.