Instant runoff, approval voting, and the status quo

(A guest post by Roberto Silvers)

After much squabbling, the Venezuelan opposition coalition has decided to hold its presidential primary elections on February 12, 2012. While this date is three months later than a majority of opposition voters and likely candidates would like, it is a fine date for the primary.

The National Electoral Council could call for presidential elections anywhere between July and December 2012. If the opposition exercises adequate planning, the selected candidate will be prepared to put up a competitive fight in national elections against current President Hugo Chávez.

The date, however, is only the beginning of several major decisions facing opposition leadership in the coming days and months. In light of this, I think the opposition should seriously consider using Instant Run-Off or Approval Voting (votación multiple) in the primary election.

Instant Runoff vs. Approval Voting

Given that the pool of opposition presidential candidates is very large as well as diverse, it is essential that the Democratic Unity Table (Mesa de la Unidad Democrática, MUD) implement a “multiple voting” (votación multiple) system for its primary election. “Multiple voting” allows voters to select more than one candidate, thereby eliminating the natural limitations and dangers of “simple voting”. Without “multiple voting”, it is likely that the opposition primary will select the “least-worst” candidate, but not the one most qualified and most likely to beat Chávez.

“Multiple voting” allows the benefits of a run-off (segunda vuelta) election without the cost of actually holding a second election. “Multiple voting” has also been endorsed by the civil society organization Súmate and Clay Shentrup of

There are essentially two “multiple voting” options: Instant Runoff Voting (IRV) and Approval Voting. IRV most closely replicates a second round election, but brings with it additional problems. Approval Voting is much simpler for the voter and easier to implement for election authorities.

Instant Runoff Voting (IRV)

  1. Each voter ranks candidates in order of preference. “1″ for my favorite candidate, “2″ for my second-favorite, etc.
  2. To calculate a winner, “1″ votes are counted first.
  3. If the winner does not receive more than 50%, “2″ votes are also counted.
  4. If the winner still does not have 50%, “3″ votes are counted.

As you can see, the IRV system is somewhat complicated to understand, and shares some drawbacks with “simple voting” systems.

Approval Voting

  1. Each voter votes for all the candidates s/he likes.
  2. The candidate receiving the most votes is the winner.

Approval Voting is much more straightforward. In addition, Approval Voting has the advantage of being able to utilize existing ballots and electronic voting machines. With IRV, on the other hand, the machines would need reprogramming to read number rankings.

In terms of voter preference, one can easily see how Approval Voting more closely represents the will of the people. Imagine two types of voters: 1) dedicated Capriles supporters who hate the other candidates; and 2) opposition voters who like Capriles, López and Machado equally. Approval Voting allows Voter #1 to vote for just one candidate while Voter #2 is able to vote for three.

The Perú Example

Imagine, for example, what the results of the Perú first-round presidential election would have looked like if Approval Voting had been utilized. Would Luis Castañeda, Pedro Pablo Kuczynski and Alejandro Toledo have cancelled each other out in the same fashion? Could Approval Voting have prevented Peruvians from being faced with the much-lamented Fujimori-Humala decision in the second-round?

Translate that to the current crop of candidates and you can see the danger for the opposition. Imagine the candidates are Capriles, Lopez, Machado, and Ledezma. In such a diverse field, there is a real danger that the vote for young moderates is split, while the votes for the dinosaurs representing the opposition’s old guard is consolidated.

In “simple voting” systems with a wide range of candidates, extremist and polarizing candidates are rewarded. With Approval Voting, consensus builders and unifying leaders are more likely to prevail. Which of these two types of leaders would best represent the Venezuelan electorate as a candidate in the 2012 presidential elections?

Roberto Silvers is a US American living, working and writing about politics, democracy and culture in Caracas, Venezuela. He has a blog and a comedy news show. Follow him on Twitter.

22 thoughts on “Instant runoff, approval voting, and the status quo

  1. This is a fine post. Thanks Roberto. Not to get too picky, but I would only add a warning on the IRV voting, that has to be with the fact that preferences are difficult (impossible?) to aggregate and with that system it is possible to end with weird results. For example if you have three voters, one voter preferences are PP>AL>HCR, another’s AL>HCR>PP and the other’s HCR>PP>AL, then PP beats AL in two of the votes (which would mean PP>AL), also AL beats HCR in two of the votes (which would mean) AL>HCR…but, this is the weird part, also it is implict that HCR beats PP (in two of the votes), so you have a three-way tie with a very weird distribution of the vote that is supposed to serve as un-tie criteria. This is called Condorcet’s Paradox btw


  2. Adoption of the AV (alternative vote) system v FPTP (first past the post) is to be define through a referendum here in Britain. AV is what got Red Ed Miliband elected as Labour Party leader. In fact, second preferences from ballots showed that Red Ed was the preferred second preference of the most radical members of the Labour Party.

    An option for Venezuela’s opposition? I wouldn’t be so sure…


  3. Omar, I agree that IRV can produce some wacky results. That’s why I emphatically recommend Approval Voting instead. With approval voting, everybody votes for everybody they like; and whoever gets the most Likes wins!

    Studies show that voters are significantly happier with results when Approval Voting is used; in other words, the will of the people is more accurately represented.

    Alek, we just need to convince four or five of the opposition leadership. Who’s ready to make a few telephone calls?


    • I’d stick to the try and tested method of one man, one vote. It works very well, provided the counting is done transparently.

      The last thing Venezuela and its fragmented opposition need is the introduction of yet more reasons for any one party to kick the table in disagreement for some complicated method of voting.

      I think yours is a harebrained idea Roberto. AV, or whatever you guys want to call it, is nothing but unnecessary trouble to increase the chances of the less abled candidates. Me thinks it’s just bullshit.

      One man, one vote. There’s absolutely nothing wrong with that system. In the UK, and other parts of Europe given the huge democratic deficit of having to put up with policies devised by officials often elected with less than a third of the roll, there may be a case for it. In Venezuela, where hasta los muertos votan, there isn’t.


  4. Omar,

    Yes, IRV has the most severe incidence of “weird results” of any commonly proposed alternative voting method.

    Condorcet’s Paradox is a result of mistakenly treating voter welfare values as ordinal, when they are in fact cardinal. If you use the Bayesian Regret metric, you avoid this trap. Approval Voting has very low Bayesian Regret – vastly better than IRV. It is also better in just about every other objectively quantifiable way.


  5. What you describe as “Instant Runoff Voting” seems to be Bucklin Voting, not IRV. Bucklin adds in second and third place votes until somebody has at least 50% of the vote, whereas IRV eliminates candidates with the least number of votes and reshuffles that candidate’s voters to other candidates based on the individual voter’s ballots.

    Bucklin is “precinct summable” and monotonic, unlike IRV. All in all Bucklin is a much more reasonable system than IRV, though I believe Approval Voting to be superior to both.


    • There was a professor out of Texas that had a software for the following method: For any set of choices, the choices would be paired in every combination, randomnizing the order. For each pair of choices, A and B, a voter would select if he preferred A over B by “a lot” or “a little”, or if he liked A and B equally, or if he preferred B over A by “a lot” or “a little”. The accuracy with which people made complex, verifiable decisions with this method was astounding.

      How would such a method rate in the Bayesian Regret scale?


      • I believe a Score voting method would be simpler for the voter and convey about the same amount of information (if not more).
        Just score each candidate between 0 to 9.


        • Scoring faired worse than paired voting when voting on items of complex or multiple factors, such as buying a car or choosing a candidate.


  6. For Multiple Voting to have any real meaning you have to include candidates that could be much more acceptable to those disillusioned with the Cuarta and Quinta. If you limit the candidates then the results and significance of Instant Runoff and Multiple Voting become exactly the same and you have only fooled yourself.

    What you are really looking for is that candidate those disillusioned with the Cuarta and Quinta would be most likely to vote for and who is still acceptable to the opposition.


  7. Doesn’t it strip away the democratic advances of holding a presidential primary if MUD limits the candidates? Even if the list were just six candidates (e.g. Capriles, Corina, Ledezma, López, Pérez and Rosales), Approval Voting would help ensure a better result. I would hate to see the young vote split, thereby allowing Ledezma or Rosales to win.

    Kepler, Leopoldo not happy? About what?


  8. After some research on the matter I find Rated voting better than Ranked voting. Approval is the simplest form of Rated voting. But I would prefer to give each candidate a score between 0 to 9 or perhaps a qualification like:

    – Very Good
    – Good
    – Somewhat good
    – Indiferent
    – Somewhat bad
    – Bad
    – Very Bad

    Within Ranked voting Condorcet is probably the best method followed by Borda. IRV is the worst.


  9. Approval voting. Me gusta la explicación que da la pagina que posteo el amigo Clay Shentrup. No se divide el voto entre A y B para que gane C… Y mientras que en la votación normal si votas por mas de un candidato seria voto nulo, con Approval voting no ocurriria esto, por lo cual se reducen los votos nulos… Ahora en lo que no estoy seguro es si incluir o no half approval…



Comments are closed.