Los Angeles Ponders Instant-Runoff Voting

This article by Lance Howland, in PublicCEO.com, explains that city officials in Los Angeles are thinking seriously about Instant-Runoff Voting. Los Angeles now uses two-round elections for its own city elections. Thanks to Rob Richie for the link.


Comments

Los Angeles Ponders Instant-Runoff Voting — No Comments

  1. IRV = THE method to elect Stalin/Hitler extremists for single office — and defeat compromise candidates.

    IRV ignores most of the data in a Place Votes Table.

    Super extreme example. Times are tough – think 2010. The *middle* is almost gone.

    49 H–W–S
    49 S–W–H
    1 W–H–S
    99

    IRV W loses, H beats S 50-49.

    First 2 place votes (Bucklin) –

    H 50, S 49, W 99

    Head to Head (Condorcet) —
    W beats H 50-49
    W beats S 50-49
    W wins.

    Approval Voting – W likely wins (as in Bucklin)

    Clones H Hitler, S Stalin, W Washington, George

    Save the example for future reference.

    It only takes ONE IRV FATAL election for some really EVIL stuff to happen.

  2. Demorep:
    Your example assumes 49% of voters support Hitler. Using this assumption, it is easy to construct similar scenarios under ANY voting system that result in “really EVIL stuff” happening.
    Really evil stuff can happen anytime voters support really evil candidates. That will never change.
    Your example therefore demonstrates nothing.
    All voting systems have problems. For single position elections, IRV runs into the fewest problems in practice.

  3. I agree that demorep’s (oft-repeated) example is a bit… over the top. But the concern is genuine. Consider this similar example, with smaller numbers:

    4 A>B>C
    2 B>A>C
    3 C>B>A

    B is preferred over any other candidate. Specifically, B is preferred over A, 5:4; and B is preferred over C, 6:3. And yet, A wins.*

    If C would have dropped out of the race, B would have won, but C is a loser regardless of who he faces; C, then, is a spoiler.

    And that’s the real problem with IRV, it doesn’t fix the spoiler problem, which is the #1 impediment to progress for 3rd parties.

    But what demorep is trying to illustrate (in an extreme way) is that it is often the candidates who fall in between the extremes that have the most to lose under IRV; in this example, B is a centrist candidate, but he’s eliminated, leaving the battle to the two extremes (and Hitler, being the most extreme example of the extreme, is his chosen standard-bearer). And the problem because exponentially worse as more candidates are added to the mix.

    I think demorep could get by with doing less Godwining, and focus on what’s important: IRV has spoilers and being a spoiler, or the fear of being a spoiler, is what kills third parties.

    After all, Australia has used IRV for years, and they haven’t elected Hitler; but they are still two-party dominated in their IRV seats.

    *Don’t read this as an endorsement of Condorcet methods; but Condorcet’s intuition does reveal the absurdity in this case. I firmly attest that approval and/or score voting are the only real solutions for single-winner elections.

  4. Oh, one more thing:

    The C-first voters in the example will realize that their LEAST favored choice won; however, if they had made the TACTICAL choice to disingenuously rank B above C, then they would have achieved a better outcome (electing B, their second choice, rather than A, their last.)

    This choice precisely mirrors the “lesser of two evils” decision-making process that 3rd party voters face now: vote honestly, and your LEAST favored candidate win; or vote tactically, and your SECOND favorite wins.

    Contrast this to approval voting. If the C-first voters know they are likely to come up short, then can approve both C and B; an honest vote that also succeeds tactically.

    Similarly, the A-first voters might not feel fully confidant of the outcome of the election in advance, and may hedge their vote for A by also approving B. Those unfamiliar with approval voting often attempt to assault it using this sort of example, saying that the A-first voters risk ruining their own best-choice. But the VERY IMPORTANT difference here is this: when making that tactical decision, the possible electoral outcomes vary between their FIRST AND SECOND choices, and NOT between their first and last as they do under plurality and under IRV.

    (And under a score voting system, voters could sort of “probabilistically hedge” their preferred vote, by offering secondary choices less than the maximum possible score.)

    This is why approval voting (or score voting) would be such a huge boon to third parties, as compared to IRVs modest (if at all) improvements. And to reiterate, the disparity only goes further in approval’s favor as we get to four or more candidates.

    Which is to say @Lems: nuh-uh.

  5. Lance Howland misrepresented what was at issue in the SF lawsuit – which was whether SF could continue to prevent voters from expressing full preferences. The judge ruled that voters who happened to prefer less popular choices would not be disenfranchised, because they could be more calculating in their voting, and figure out who was likely to be in contention, so that their ballot would not be discarded. Voters in Alameda County should demand a comprehensive education program that teaches them how to cynically game an IRV election.

    He also continues to suggest that Arnold Schwarzenegger should have delayed the special election in SD 15 until November. Based on how the Democrats are tooting this issue, it becomes more evident that they deliberately delayed confirming Abel Maldonado as Lieutenant Governor. Had Maldonado been confirmed one week earlier, the special primary could have been coincident with the general primary on June 8. Delaying one week, combined with using 9 of the 14 days in which the governor has to call a special election would barely permit a special general election in November.

  6. STOP IRV — at just about ALL COSTS.

    These are EXTREME times with EXTREME candidates for major offices.

    P.R. legislative

    A.V. executive/judicial — to elect those dull middle folks to be in jobs like Prez, Guv, local Hiz Honor [Mayors], etc.
    ——–
    #3 fter all, Australia has used IRV for years, and they haven’t elected Hitler; but they are still two-party dominated in their IRV seats.

    And what if IRV DOES elect a Hitler clone in Aussie land ??? A new base for nazi world conquest ???
    ——-
    The standard *large middle* divided example is —

    34 H–W–S
    33 S–W–H
    16 W–H–S
    16 W–S–W
    99

    Does the U.S.A. have *at least* 67 percent of the population in the left/right extremes (since about 1929) ???

    Does the Sun rise in the East and set in the West ???

  7. As to the quality of the candidates elected*, on average IRV is, in the worst case, as bad as plurality. But it is not _worse_ than plurality.

    If you’re afraid IRV will elect Hitler, then you should have a greater fear of Hitler getting elected today.

    *This doesn’t account for non-outcome based measures (such as counting difficulties or voter confusion).

  8. I have been in favor of IRV for a long time, I had problems with it. In some contexts, you cannot check the vote count, either by a full recount, nor by a spot check. Some of those contexts are:

    1) in contests spanning several counties, in order to check the vote, you would have to coordinate the recount by the counties, because you would not want to proceed with a check of a following round until you knew who won/lost the earlier rounds. The reality is that each of California’s 58 counties conducts elections according to their own procedures. Trying to get them to coordinate their counting would be difficult at best. Doing so for all 58 counties in a statewide election would be a nightmare. My conclusion here is that it is not practical to audit or recount the vote of an IRV election that spans several counties, and definitely not statewide.

    2) the IRV version of voting for several candidates to fill several seats on, say a school board, is called Single Transferable Vote (STV). As I learning in observing last fall’s KPFA (Berkeley radio) election, the normal procedures as you work through the rounds is to divide votes into fractions. I was able to read through a computer report of how a program calculated carried the fractional votes from round to round, when there were a few seats to fill, and less than 200 ballots. I was satisfied the the counting program was working properly in that case. When it came to the larger election, with roughly 3000 ballots, and more seat, the checking became very difficult because keeping track of many fractional votes was too much. I came to the conclusion that an STV election is not practically hand-checkable when there are more than somewhere between 1000 and 10,000 ballots. It should not be used in an election for public office in California, because recounts are not workable.

    3) The statement was made by an IRV advocate that IRV elections are auditable because images of the ballots are posted online. I talked with him afterwards and attempted to correct him, as California does not post images of all ballots online. He insisted that it does. I asked him to send me a link proving that claim. That was 2 weeks ago. I have heard nothing. I stand by my conclusions.

    4) I have been involved in the election integrity movement for 5 years now. I have seen a significant number of experts say that you cannot spot check (aka audit) an IRV vote, because it’s too complicated with the multiple rounds. I think you can, within a single county, and several of us on the San Francisco Voting Systems Task Force will be reviewing how to do it. I expect to announce the results of that review on my website when it is completed later this year.

    Jim Soper

    Co-Chair, Voting Rights Task Force (SF East Bay)
    Member, San Francisco Voting Systems Task Force
    Author of http://www.CountedAsCast.com
    Senior software consultant

  9. #8 (item 3). You (and perhaps your correspondent) may be confused with what is meant by “ballot image”. It is not a digitized photographic image of the ballot face, but rather simply a reduction of each ballot to its preferences.

    If you go to the SF Elections Commission website http://www.sfgov2.org/index.aspx?page=599

    and click on “Election Results By Year”, “2008”, “Election Summary (for November 4, 2008)” you will see “City Wide RCV Ballot Image Report (ZIP)” The zip file includes a PDF file describing the format for the image files. In this case, they are for the 2008 Supervisors elections.

    Assuming that you use paper ballots (including optically scanned ballots), it is straightforward to hand sort them by preference order. You start out by sorting ballots into stacks based on their preference order, just as you would do in a hand count of a FPTP election. After you finish sorting, you count the ballots. You then repeat the process with each stack of ballots with a common first preference. After that step you will have a stack for each permutation of first and second preference. And so on.

    The ballot image file which was presumably generated by a machine scanning operation can be processed in an analogous fashion, and then the two can be compared.

    In Minneapolis, the ballots were hand sorted on a precinct by precinct basis. This was done at a central counting center after the election, but there is not a particular reason that this could not be done on election night at the polling place.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.