It certainly is interesting. It proposes a notion that I'm sure didn't only occur to you and me, and shows that that notion is probably false. Thereby improving our understanding of the world. Thank you.
Wouldn't it make sense to account for electoral votes in the graph somehow? Plotting HI with the same weight as CA seems a little silly. I am not really sure how best to do this, perhaps size of the data point? It may simply be that there is a correlation but only among the larger more important for an election states. I do not really care what happened in Arkansas.
You graph also appears to fit a linear relationship pretty well. I bet a regression on the data would give you an R*R of over .5 though the line has almost no slope. This suggests to me that the voters just pretty much uniformly preferred Obama over McCain 10% more than they did Kerry over Bush.
It's normalized because it's done in percent, not in numbers of votes.
In theory, to increase voter turnout by some percent in a state with 1/2 the population of another state, one would only have to devote half the resources to the former as to the latter.
Right - but due to dramatic efficiencies of scale that is certainly not the case. I agree with your basic premise though.
It would be interesting to see how well correlated the GOTV effort was with regard to Obama swing in those states where Obama actually tried to GOTV. People in Arkansas heard the news about GOTV and there is no doubt in my mind that effected their turnout too. No amount of GOTV was gonna swing Arkansas though.
registering voters, which takes people roughly in proportion to population to set up booths, knock on doors, whatever.
Finding out voter preferences, which takes people calling roughly in proportion to the population.
On election day, getting 'your' voters to the polls by checking at each precinct who among 'your' votes has not yet voted and calling them repeatedly, trying to get them to vote, so proportional to number of precincts (i.e., population)
Even advertising is, I assume, basically $/viewer.
Comments 10
Reply
You graph also appears to fit a linear relationship pretty well. I bet a regression on the data would give you an R*R of over .5 though the line has almost no slope. This suggests to me that the voters just pretty much uniformly preferred Obama over McCain 10% more than they did Kerry over Bush.
Reply
of votes.
In theory, to increase voter turnout by some percent in a state
with 1/2 the population of another state, one would only have
to devote half the resources to the former as to the latter.
Reply
It would be interesting to see how well correlated the GOTV effort was with regard to Obama swing in those states where Obama actually tried to GOTV. People in Arkansas heard the news about GOTV and there is no doubt in my mind that effected their turnout too. No amount of GOTV was gonna swing Arkansas though.
Reply
There's three stages:
registering voters, which takes people roughly in proportion to population to set up booths, knock on doors, whatever.
Finding out voter preferences, which takes people calling roughly
in proportion to the population.
On election day, getting 'your' voters to the polls by checking at
each precinct who among 'your' votes has not yet voted and calling
them repeatedly, trying to get them to vote, so proportional to
number of precincts (i.e., population)
Even advertising is, I assume, basically $/viewer.
So where are the dramatic efficiences of scale?
Reply
Leave a comment