The Otherworldly Circus - The America Thread

[hujsh]

I'm talking about the presidential campaign. Congressional seats is different to the presidential race.

Yes, my point is that is the only incentive for presidential candidates to bother campaigning in those safe states in the EC system. Because they want a majority in congress to enact their agenda.

Under a popular vote system while there will be more incentive to campaign in places with more people, each part of the country will still have a say and if it looks like being close you'll want to appeal to as many people as possible (so also the small states). There will always be a reason to vote.

 

[Norm De Guerre]

Remember kids if you want to succeed at school.

Vaccines are bad, but street heroin is ****in' great at boosting your grades!



Theres always a video with these clowns from a time before they red pilled themselves.

So here's another one.

 

[yebach]

Remember kids if you want to succeed at school.

Vaccines are bad, but street heroin is ****in' great at boosting your grades!



Theres always a video with these clowns from a time before they red pilled themselves.

So here's another one.

Footscray was once the Heroin capital of Australia.
Could explain why we've got so many intelligent and eloquent posters on this board.

 

[Dogs_R_Us]

To be fair, that's how elections work though. In Aus, if Vic and NSW overwhelmingly voted one party, the rest of the country wouldn't be able to influence the result. In Canada its Quebec and Ontario.
Even our preferential system here is flawed, youve got aome candidates getting a HoR seat, despite getting lower primary votes than other candidates, but get in purely due to preference deals.
The only true and fair system is first past the post.

In the preferential system the majority don’t get who they don’t want.

 

[Arkangel]

It might favour the major parties, but first past the post is literally just 'majority rules'. To me that seems the fairest, but happy to hear/read other perspectives.
It might favour major parties at an operating government level, but there are many seats where minor parties get better vote tallies than the major parties.

As I understand it, first past the post (FPP) is only "majority rules" when there are only two candidates to choose between. In a single-transferrable vote (STV) system like we have, if there are only two options the system becomes FPP by default. Further, if there are more than two candidates but one of them is preferred over all others by a majority of the voters, they would win under both systems. So, in these scenarios there can be no superiority of one approach over the other.

Where the two systems diverge is cases where no candidate receives a majority of the votes. In FPP the winning candidate is then the one with the greatest number of votes, irrespective of how small that number is as a proportion of the total votes. In STV, the winning candidate is basically the one that the majority of voters prefer relative to the remaining options (a bit of a simplification, but basically what it amounts to).

Which system does a better job of electing candidates who reflect voters' preferences? My intuition was that STV would do a better job of this, but since I don't like trusting intuition I ran a simple simulation to try to answer the question. In the simulation, I assumed an electorate of 10000 voters whose political preferences vary across two dimensions according to independent normal distributions (you can think of these as economic and social dimensions, like the political compass, if that helps). This leads to a distribution of voters' preferences that looks like this (where each blue point is a single voter):


I then simulated the positions of some number of political candidates; the figure above shows an example with 3 candidates (the red dots). For each voter, I calculated the straight-line distance between their own preferences and those of the candidates. Under FPP, I assumed that each voter selects the candidate they are closest to in this space, whereas for STV I assumed each voter preferences the candidates in ascending order of distance (i.e., first preference is the closest candidate, second preference is the next-closest, and so on).

I ran this simulation 10000 times for each of FPP and STV, and then calculated the average distance across all voters from the winning candidate's position. The number of "victories" for each electoral system from these 10000 simulations, when varying the number of political candidates, was as follows:

Number of candidates	FPP wins (of 10000)	STV wins (of 10000)
3	4792	5208
4	4530	5470
5	4267	5740
6	3938	6064
7	3724	6285
8	3494	6515
9	3423	6585
10	3141	6870

Note that some rows sum to slightly > 10000 because of ties or NAs in the simulated data (I didn't spent the time to totally iron out all the kinks in the code). In any case, this pretty clearly shows that STV is always superior to FPP by this metric, and that this superiority increases with the number of candidates in the electorate.

My disclaimer here is that I'm not an expert in political science, so I may be missing the knowledge that would allow me to identify important features not captured here (e.g., I don't know if there's research out there about the distribution and dimensionality of voters' preferences, so assuming it's 2-dimensional, independent, and normal may be way off). This was just my simple but systematic way of working through the problem to test my intuition.

If anyone wants to re-run or modify the simulation, the R code for the simulation function is below; it returns a vector of distances of voters' positions from the winning candidate's/party's position under the specified electoral system.

runSim <- function(nVoters, nParties, choiceRule=c("STV", "FPP")){
partyPositions <- data.frame(party=1:nParties,
dim1Pos=rnorm(nParties),
dim2Pos=rnorm(nParties))

allPrefs <- data.frame(voterN=1:nVoters,
dim1Pos=rnorm(nVoters),
dim2Pos=rnorm(nVoters))

for (i in 4:(nParties+3)){
allPrefs[,i] <- sqrt((allPrefs$dim1Pos-partyPositions$dim1Pos[i-3])^2 +
(allPrefs$dim2Pos-partyPositions$dim2Pos[i-3])^2)
}

voterN <- rep(0, nParties)
excluded <- rep(0, nParties-1)

if (choiceRule=="FPP"){
for (i in 1:nParties){
voterN[i] <- sum(ifelse(allPrefs[,i+3]==do.call(pmin, allPrefs[-c(1:3)]), 1, 0))
}
distFromWinner <- allPrefs[,which(voterN==max(voterN))+3]
return(distFromWinner)
} else {

for (i in 1:nParties){
voterN[i] <- sum(ifelse(allPrefs[,i+3]==do.call(pmin, allPrefs[-c(1:3)]), 1, 0))
}

for (i in 1:(nParties-1)){
excluded[i] <- which(voterN==(sort(voterN)[i]))

for (j in 1:nParties){
voterN[j] <- sum(ifelse(allPrefs[,j+3]==do.call(pmin, allPrefs[-c(1:3, excluded+3)]), 1, 0))
}

}

distFromWinner <- allPrefs[,which(voterN==max(voterN))+3]
return(distFromWinner)

}
}

 

[dogwatch]

As I understand it, first past the post (FPP) is only "majority rules" when there are only two candidates to choose between. In a single-transferrable vote (STV) system like we have, if there are only two options the system becomes FPP by default. Further, if there are more than two candidates but one of them is preferred over all others by a majority of the voters, they would win under both systems. So, in these scenarios there can be no superiority of one approach over the other.

Where the two systems diverge is cases where no candidate receives a majority of the votes. In FPP the winning candidate is then the one with the greatest number of votes, irrespective of how small that number is as a proportion of the total votes. In STV, the winning candidate is basically the one that the majority of voters prefer relative to the remaining options (a bit of a simplification, but basically what it amounts to).

Which system does a better job of electing candidates who reflect voters' preferences? My intuition was that STV would do a better job of this, but since I don't like trusting intuition I ran a simple simulation to try to answer the question. In the simulation, I assumed an electorate of 10000 voters whose political preferences vary across two dimensions according to independent normal distributions (you can think of these as economic and social dimensions, like the political compass, if that helps). This leads to a distribution of voters' preferences that looks like this (where each blue point is a single voter):

View attachment 2176139
I then simulated the positions of some number of political candidates; the figure above shows an example with 3 candidates (the red dots). For each voter, I calculated the straight-line distance between their own preferences and those of the candidates. Under FPP, I assumed that each voter selects the candidate they are closest to in this space, whereas for STV I assumed each voter preferences the candidates in ascending order of distance (i.e., first preference is the closest candidate, second preference is the next-closest, and so on).

I ran this simulation 10000 times for each of FPP and STV, and then calculated the average distance across all voters from the winning candidate's position. The number of "victories" for each electoral system from these 10000 simulations, when varying the number of political candidates, was as follows:

Number of candidates	FPP wins (of 10000)	STV wins (of 10000)
3	4792	5208
4	4530	5470
5	4267	5740
6	3938	6064
7	3724	6285
8	3494	6515
9	3423	6585
10	3141	6870

Note that some rows sum to slightly > 10000 because of ties or NAs in the simulated data (I didn't spent the time to totally iron out all the kinks in the code). In any case, this pretty clearly shows that STV is always superior to FPP by this metric, and that this superiority increases with the number of candidates in the electorate.

My disclaimer here is that I'm not an expert in political science, so I may be missing the knowledge that would allow me to identify important features not captured here (e.g., I don't know if there's research out there about the distribution and dimensionality of voters' preferences, so assuming it's 2-dimensional, independent, and normal may be way off). This was just my simple but systematic way of working through the problem to test my intuition.

If anyone wants to re-run or modify the simulation, the R code for the simulation function is below; it returns a vector of distances of voters' positions from the winning candidate's/party's position under the specified electoral system.

runSim <- function(nVoters, nParties, choiceRule=c("STV", "FPP")){
partyPositions <- data.frame(party=1:nParties,
dim1Pos=rnorm(nParties),
dim2Pos=rnorm(nParties))

allPrefs <- data.frame(voterN=1:nVoters,
dim1Pos=rnorm(nVoters),
dim2Pos=rnorm(nVoters))

for (i in 4:(nParties+3)){
allPrefs[,i] <- sqrt((allPrefs$dim1Pos-partyPositions$dim1Pos[i-3])^2 +
(allPrefs$dim2Pos-partyPositions$dim2Pos[i-3])^2)
}

voterN <- rep(0, nParties)
excluded <- rep(0, nParties-1)

if (choiceRule=="FPP"){
for (i in 1:nParties){
voterN[i] <- sum(ifelse(allPrefs[,i+3]==do.call(pmin, allPrefs[-c(1:3)]), 1, 0))
}
distFromWinner <- allPrefs[,which(voterN==max(voterN))+3]
return(distFromWinner)
} else {

for (i in 1:nParties){
voterN[i] <- sum(ifelse(allPrefs[,i+3]==do.call(pmin, allPrefs[-c(1:3)]), 1, 0))
}

for (i in 1:(nParties-1)){
excluded[i] <- which(voterN==(sort(voterN)[i]))

for (j in 1:nParties){
voterN[j] <- sum(ifelse(allPrefs[,j+3]==do.call(pmin, allPrefs[-c(1:3, excluded+3)]), 1, 0))
}

}

distFromWinner <- allPrefs[,which(voterN==max(voterN))+3]
return(distFromWinner)

}
}

This is great. As you warn, it is a simplified model and might not totally reflect reality (I can think of a few examples) but it mounts a compelling starting position for STV being a better system.

It also shows how a FPP system can become less satisfactory the more candidates there are. This in turn suggests how it might be undermined by fielding extra candidates (possibly stooges) who can draw votes away from the presumptive top contender.

That's not to say that STV systems can't be gamed too, but in Australia I think that has only really happened in multi-member electorates. The more members to be elected the easier it is to get a candidate up with a low first preference count. It has occurred once or twice in the Senate where some obscure candidates have got up Bradbury-fashion with a ridiculously low first preference count after using a preference whisperer. I think there have been tweaks in recent times to minimise the chances of that happening again. It doesn't appear to have been an issue in single member electorates which is what we have in the House of Reps.