VBM and the costs of elections: updates from CO, IN, and WA

Changes to early voting are on the administrative and legislative calendar in many states and jurisdictions.  A brief update from the mail bag:

  1. In Colorado, many counties have absentee ballot rates exceeding 70%.  This was the threshold, by the way, that encouraged Oregon and Washington to move to full vote by mail.  In Arapahoe County, less than 15% of the electorate is showing on up election day, and the county is contemplating a substantial reduction in precinct places, using election day vote centers instead.  Officials estimate that they can reduce the number of election judges by 2/3.
  2. County clerk Kathy Neal, in Summit County IN, is proposing that all primaries and elections held in odd number years be conducted completely by mail.  Neal expects this to result in a 25-30% cost savings by eliminating the need for election judges. Already , 40% of ballots come in via the mail in Sumit County.
  3. Pierce County, WA may finally have to move to full vote by mail elections, if a bill currently moving through the legislature passes and is signed by the governor.  Part of the argument is fiscal: county auditor Julie Anderson estimates that the county spent $16.97 per precinct place vote (10,000 in 2010) vs. $6.88 per vote by mail ballot (135,000).

Comments 1

  1. Martin wrote:

    Anthony,Before I even get into the details of the eaxmple election I cited, I want to point out that the realism of that eaxmple is completely irrelevant to the point I’m making. The point is that a weak party is statistically more likely to defeat ONE of the major parties than to defeat BOTH of them. This should be obvious. It’s the same concept as if you’re in a race with two Olympic track athletes. You’re more likely to defeat ONE of them than to defeat BOTH of them. (And even MORE likely to defeat neither of them.)”AV makes it more difficult to determine which two candidates are most likely to win, compared with FPTP”Not really. If you don’t have a strong indicator from pre-election polls, you can just go off historical data. If you live in an area where the winner is usually a Tory or Labour candidate, you just KNOW that, statistically speaking, one of them is most likely to win.In Australia’s House of Representatives, for eaxmple, the winner is in a party other than Labor or NatLib only about 0.1% of the time. I have also gone through decades of Burlington history, and I cannot even FIND a mayor that I can confirm was a Republican. It is obvious to locals that the Democrat and Progressive candidates are clearly the strongest two.On that, note the Republican was NOT a front-runner — he was the Condorcet LOSER among the final three candidates. He would have lost against the Democrat even WORSE than he lost against the Progressive. He only LOOKED strong because the liberal candidates split the first-place rankings of liberals.And the Democrat was preferred by a majority to the winner as well (so much for the myth that common pro-AV claim that it guarantees majority winners). The Democrat was the Condorcet winner (and apparently the social utility maximizer, or “right winner”), but the Progressive won due to IRV’s completely ignoring the Republicans’ preference of Democrat over Progressive. IRV literally ignored that information, since it only looks at one layer at a time.ScoreVoting.net/IgnoreExec.html”If I don’t particularly like either M1 or M2 then my strategy is to vote for m, and that is always the best strategy, in order to maximise the chances of m winning.”Only in EXTREMELY rare circumstances in which your utility difference for M1 and M2 is virtually zero. Consider this hypothetical expected utility calculation:Your utilities:U(m)=100, U(M1)=1, U(M2)=0In other words, you prefer m over M1 by 99 times as much as you prefer M1 over M2.Now say that swapping your sincere vote for a first-ranking of M1 has:A 0.0001 chance of changing the winner from M2 to M1.A 0.000001 chance of causing M1 to win instead of m.So your expected utility by making that move would be:(0.0001 * 1) + (0.000001 * -99) = 0.000001 = a positive expected valueYou can plug in any numbers you want, but for the vast majority of people, any remotely realistic values will support the insincere vote.Anyway, back to my eaxmple. It essentially used M1=Progressive, M2=Democrat, m=Republican, to approximate Burlington.It would also work with M1=Republican, M2=Democrat, m=Green, but then you might want to turn some of the Dem>Rep>Green voters into Dem>Green>Rep voters, to make it more realistic. But then it also starts to get harder for people to follow, and I thought it best to keep it simple. It’s always a trade-off when trying to explain these things to a lay audience, but in a way that politicos will still find “realistic”.

    Posted 30 Mar 2012 at 7:11 pm