F10 Tweaking the model - MCMC and types of elections Flashcards
What is the function of endorsements?
Voters can ‘borrow’ a position from a referent person. Aimed at independents
What are the five types of elections?
- Maintaining dominance
- Deviating from dominance
- Re-instating dominance
- Balancing
- Realignment of groups
1-3 happens when there is a clear majority in the normal vote. 4-5 when the normal vote is balanced.
What is the normal vote?
Distribution of party identification between dem and rep discounted by differences in turnout patterns of supporters of each part.
What is a ‘maintaining election’?
Stable partisan attachments continue to be a major determinant of election result
What is a ‘deviating election’?
Short-term partisan attitudes deviations lead to election of presidential nominee of the minority party
What is a ‘reinstating election’?
The majority party wins the White House back after a deviating election
What is a ‘realigning election’?
Majority party of the normal vote loses election after election because the partisanship of people changes (long term)
Not necessarily unidirectional.
What is a ‘balancing election’?
Neither party has a majority in party identification.
What are three examples of a realignment that political scientist agree on?
Social groups are basis of a realignment.
1860 (after the civil war - Lincoln and anti-slavery)
1896 (shift in regional basis - Republicans strong in the Northeast and Midwest, and Democrats in the South and rural areas)
1930 (the New Deal era - Roosevelt after 30’s depression. Social welfare policies uniting ethinic minorities, urban and poor)
Which elections have been balancing?
All elections since 1980 more or less since scientist can’t agree on a realignment.
What are examples of maintaining, deviating and re-instating dominance?
Maintaining: Lyndon B. Johnson (dem)
Deviating: Eisenhower after 20 years of dem
Re-instating: JFK after Eisenhower
What is a possible realignment currently?
Education
Gender gap exploding
Race less important (Latino vote for men especially)
Why do we use Markov Chain Monte Carlo?
Because the integral for Bayes rule is to complex to solve (the parameter space is massive).
How can we randomly draw samples from a posterior distribution that is unknown?
We take a random walk through the parameter space, favoring parameter values which have a high posterior probability.
Proposed jumps in the walk are accepted/rejected probabilistically.
We end up with sampling where probability is higher thus converging the true posterior distribution.