My worry about the polls is the wording of the questions. For example, suppose you are interested in what determines whether a person supports or opposes "Trump's handling of immigration”? Responses on a Likert scale:
1. Strongly Agree.
2. Agree
3 .Disagree
4. Strongly Disagree.
( 0 =Not applicable, 8 Don’t Know, 9 Not Answered are all coded as missing values.)
Should this be modeled as an OLS model where the dependent variable is a continuous scale indicating increasing disagreement, from 1 to 4?
Or as an OLS/ Linear Probability Model for the probability of Agree vs Disagree, where Strongly Agree is collapsed into Agree, and Strongly Disagree is collapsed into Disagree?
Or as a Logistic Regression for the probability of Agree vs Disagree, with the same collapsing?
All in all, garbage in, garbage out is my view on polls.
The most important fact about the electoral college is that all states except smallish Nebraska and Maine maximize their power by casting 100% of the electoral college for the popular vote winner. This could be characterized many ways but not as "anti-majoritarian," as it over-rewards the majority in the state.
Presumably, states would also maximize their power under a popular vote system by maximizing the vote differential in their state between the majority party's presidential candidate and the alternatives. How they might do this we already know since the end of Reconstruction and recent moves to disenfranchise the opposition. With the electoral college, such moves have a limited impact, only being important in large competitive states, especially the places most able to combat the problem -- Florida in 2000 only gives you an appetizer of the problems to come with a never-tested popular vote system.
My worry about the polls is the wording of the questions. For example, suppose you are interested in what determines whether a person supports or opposes "Trump's handling of immigration”? Responses on a Likert scale:
1. Strongly Agree.
2. Agree
3 .Disagree
4. Strongly Disagree.
( 0 =Not applicable, 8 Don’t Know, 9 Not Answered are all coded as missing values.)
Should this be modeled as an OLS model where the dependent variable is a continuous scale indicating increasing disagreement, from 1 to 4?
Or as an OLS/ Linear Probability Model for the probability of Agree vs Disagree, where Strongly Agree is collapsed into Agree, and Strongly Disagree is collapsed into Disagree?
Or as a Logistic Regression for the probability of Agree vs Disagree, with the same collapsing?
All in all, garbage in, garbage out is my view on polls.
The most important fact about the electoral college is that all states except smallish Nebraska and Maine maximize their power by casting 100% of the electoral college for the popular vote winner. This could be characterized many ways but not as "anti-majoritarian," as it over-rewards the majority in the state.
Presumably, states would also maximize their power under a popular vote system by maximizing the vote differential in their state between the majority party's presidential candidate and the alternatives. How they might do this we already know since the end of Reconstruction and recent moves to disenfranchise the opposition. With the electoral college, such moves have a limited impact, only being important in large competitive states, especially the places most able to combat the problem -- Florida in 2000 only gives you an appetizer of the problems to come with a never-tested popular vote system.