On Mon, 11 Oct 2004 20:01:10 -0400, Erik Reuter <[EMAIL PROTECTED]> wrote: > On Mon, Oct 11, 2004 at 04:22:16PM -0400, Bryon Daly wrote: > > > My main problem with the system as it is now, I suppose, is that if > > you're a Kerry supporter in, say, Alabama or a Bush supporter in > > Massachusetts, your vote has zero potential for impact on the race, > > *even if the race is extremely close from a nationwide perspective*. > > I assume you mean to imply "close" between the leading two candidates. > > But what makes your "race is extremely close from a nationwide > perspective" between the leading two candidates criterion special? There > are more than 2 people running, but the race is not close between, > eg., Nader and Bush.
I didn't really see it as a criterion so much as an attempt to highlight the disparity. The point I'm trying to make is that while in some places, a few handfuls of votes may decide all the electors for the state, in other locations, potentially millions of votes for the candidate who lost that state won't have any effect on the state's electors or the national outcome - even though it's for the same election, voting for the same candidates, at both locations. Now that I think about it further, I think part of my communication problem is that I've been muddling together a few different concerns and comparing apples and oranges in terms of trying to compare/contrast "voter impact", in a way.. I'll try to sort it out better below. > If you change your criterion to make it more equitable, then it would be > something like, in an N candidate race, each candidate polls close to > 1/N of the votes. > > > I don't mean this in some nihilistic "it just doesn't matter" kind > > of way. What I mean is that a vote for Bush or Kerry in Iowa or PA > > is infinitely more important that one in MA/AL. I'd prefer if every > > person's presidential vote was of equal value regardless of which > > state they lived in. > > A vote for Bush or Kerry is infinitely more important than one for > Nader. WOuld you prefer every person's vote was of equal value > regardless of which candidate whose views they share? Perhaps yes, if by that you mean that I'd prefer a vote is a vote is a vote, regardless of where you're voting from or who you're voting for. Or in other words, if we used the straight nationwide popular vote to determine the presidential election so that per-state results did not mask out or quantize the results. Or if we used my second alternative, there'd be at least a chance for Nader or other third party candidates to get an electoral vote. > > The root cause of that problem, as I see it, is that most states award > > the winner of the state's popular vote all of the state's electoral > > vote. That means that whether only one person voted for the loser or > > 49.9% or the people did, its essentially treated as if 100% voted for > > the winner as far as the EC is concerned. > > I don't think that is the root cause of your objection. If you had > a nationwide popular election (no EC), and pollsters surveyed a > representative sample of 100,000 people just before the election, their > survey would have only a 0.3% margin of error. If it came out, say, > 50.5% to 49.5%, then everyone voting for the 49.5% candidate would, in > high likelihood, be voting for the loser. You could say that the problem > is that in a winner takes all election that 49.5% of the votes are > wasted unless the race is closer than 0.3%. Which doesn't make a lot of > sense to me, but that is very similar to what you are saying above. There is a difference (or at least I perceive one). My point above is actually based on the same objection/concern as the worst-case scenario I mentioned. Let me try to explain my thoughts in a (hopefully!) less muddled way... As I see it, the state "winner takes all EV's" unnecessarily forces a winner/losers to be declared on the per-state level. Doing so throws away all the voter selection information *prematurely*. Sure, in the end of any election, you have to gather all the votes, pick a winner and say "tough luck" to the 49.5% (or whatever) loser(s) but that shouldn't be done until the highest levels of vote collection. I see the electoral college as acting essentially as proxies for the will of the voters. When a state gives all its EV's to its winner, that essentially introduces an error factor where those EV's may only represent 50% of the state's voters. It's almost like a rounding error: the winner gets "rounded up", while the losers get "rounded down". Looking at CA with almost 34 million residents, let's pretend that Kerry wins there with 50% of the vote to Bush's 42% to Nader's 8%. Kerry then gets 55 electoral votes ostensibly representing CA's 34 million residents but really only representing 50% of the voters. If we can assume that the voters generally mirror the population's politics - that's in essence 17 million people being represented by the wrong guy. That's the equivalent population of the smallest 15 states plus DC! Now multiply this error factor times 50 states, and you get a heck of a lot of non-representation, and even though some of it will cancel out, there's still more than enough to allow the loser of the popular vote to win the electoral vote. It's like if you're doing a long math problem, and at every step you were forced to round your numbers to integer values: at the end you'll have an approximation, but it's not going to necessarily be that accurate. Anyway, that's why I connect my worst-case "up to nearly 75%" situation to my "winner takes all EV's" concern. So if that was my big concern, why the heck was I blabbing about blowout states vs battleground states? If anything, the tightly contested states will have closer results and so *more* people will lose out in winner takes all, compared to in the blowout states! Well, yes. That's where I muddled things up because I was trying to contrast things in vague terms of voter impact. In the battle ground states, while the losing party's voters don't get any electoral representation just as elsewhere, I was thinking that at least they had their real in-state contest that they had a shot at having some impact on, whereas in a blowout state, they don't even get that. And then of course there's the issue that the blowout states don't get any attention. But really, I see both of those as effects of the "winner takes all EV's" cause. Anyway, now imagine instead that the states award their EV's proportionally to the state's popular vote. There would still be a rounding error factor because the pop vote percentages wouldn't always align well with the state's integer EV count*, but it'd still represent the voter's will a heck of a lot better than with the winner takes all system. And so this way, we might still end up with a 49.5% loser, but at least each candidate's electors would more closely match the people who voted for him. > > It seems to me that with this system, a worst case scenario could > > result in up to nearly 75% of the voters voting for the candidate that > > loses. That's extremely unlikely, of course, but a possibility that > > illustrates the flaw, I think. > > I think this objection makes more sense than your "root cause" above. I > didn't check if your 75% number is correct, but if it is, I would > challenge anyone to come up with a rationale for how an election would > be "fair" if 75% of the votes cast went to a losing candidate. The 75% number was just quick reasoning, so maybe I got it wrong: For a worst case, let's say that 49.9% of the electoral votes will be going to the loser. Then we'll say that 100% of the voters in the loser's states voted for him. Now, let's say that all of the winner's states, he won by a .1% margin, so 50.1% voted for the winner and 49.9% voted for the loser, in the winner's states. Now sum those up: The loser's got 100% the votes for his (almost) half of the EV's (so 50% of the total vote here), plus he got (almost) half of the votes in the winner's 50.1% EV's, (so another 25% of the total vote here). For a potential total very near 75%. This is of course ignores the actual uneven distribution of state's electoral votes, or potential variations in actual voters per EV per state, but I don't think it's a terribly bad rough estimate. And of course, if you start thinking about potential worst cases with a strong 3rd party showing (say 20% across the nation, with no states won), the electoral winner could possibly receive far less than 25% of the total vote. > However, as you say, it would be very unlikely. To improve your > criterion, we probably need to quantify it. Something along the lines > of "the expected unfairness index" of an election system, which would > (roughly) be the sum of > > (probability_i) x (unfairness_i) > > for all possible unfair outcomes. Probability can be > calculated, but "unfairness" would need to be quantified in > some manner, and offhand I don't have a simple way (unless 1 - > probability_of_casting_the_deciding_vote summed over all voters would > work). Although this line of thought strikes me as rather obvious -- I > would bet that a number of political scientists have written papers on > something along these lines. Interesting. I'd define fairness as the level of ability of all voters to have equal voting power on the final election result. So by that standard, I would consider a nationwide direct popular vote to elect the president to be 100% "fair". There could still be a 49.5%loser, and/or Nader might only get 2% of the popular vote, but everyone's vote would count the same. The things that introduce unfairness IMHO are the winner-takes-all "rounding error" factor that nullifies votes before they can affect the final score, and to a lesser extent the way that electoral votes vary per capita per state. > > 2) get all the states to abandon the "winner takes all" EC vote > > allocation and go to something that allocates votes in a way more > > proportional to the popular vote. (Either a straight proportion, or > > the by-district thing). > > In this paragraph, I am going to guesstimate some numbers, but I'd > welcome any more accurate and/or supported numbers anyone can supply if > I am far off. > > Currently, I think the pollsters survey around 900 - 3000 people per > state per week. This results in a margin of error of 1.8% to 3.3% in > estimating the way the state will vote. So, if you live in a small state > that is polling, say, 52% to 48%, then most people will feel that their > vote "counts" because it is not possible to confidently predict the > outcome of the state without surveying a lot more people. > > In contrast, if we adopted your proposal (2), the outcome will be very > similar to your proposal (1) where the nationwide popular vote decides > the election (there is a slight difference, as you say, because low > population states get a little extra ooomph, but the difference is > relatively small). Imagine then that these pollsters poll 100,000 people > randomly selected across the country, which is the equivalent number > of people as 2000 x 50 states. The result is a margin of error of only > 0.3%. As I pointed out above, in a 50.5% to 49.5% race with a 0.3% > margin of error, it is very likely that the people voting for the 49.5% > candidate are voting for the loser. So you have 49.5% of the people > casting a "non-decisive vote". > > It is not at all obvious that the number of people casting "non-decisive > votes" will be less, in the average election, in your proposal 2) as > compared to the current Electoral College system. For example, if > it frequently turns out that the nationwide popular vote polls at > 50.5%/49.5% or more, while many of the most populous states poll at > about 51.5%/48.5% and 48.5%/51.5%, then it may be that the current > system is better at minimizing the number of "non-decisive votes" than > your proposed system. I'm really concerned more about non-representative electoral votes than non-decisive popular votes. If the electoral votes were proportional to the state pop vote, there would be no non-decisive votes at the state level, because there'd be no decision-making at the state level. > > I hope that clarifies things! > > Somewhat, but I think the key assumption is what "maximizes the > fairness" or "minimizes the unfairness" of an election? It seems to me > that the answer is not at all obvious, and I'm still not certain of the > criteria you use to judge whether an election is fair. I know that you > consider it unfair if 75% of the people vote for the losing candidate, > but is that your only criterion for unfairness? I think I've covered all this above. -bryon _______________________________________________ http://www.mccmedia.com/mailman/listinfo/brin-l
