by Michael Ossipoff
A number of organizations, and even a few political parties, use or advocate a voting system called Instant Runoff Voting (IRV). IRV would be suitable if everyone agreed on the definition of “acceptable”, i.e. if there weren’t any voters prone to over-compromise, favorite-burial and unquestioning giveaway to a “lesser-evil” and, finally, if there weren’t a count-fraud problem. However, as none of the aforementioned statements are true, IRV is not a suitable choice for political elections.
Under existing conditions, the following issues present some serious problems of IRV:
- Because IRV is a rank method, it is labor and computation-intensive (this is true of all rank methods). Such methods will almost surely be counted by computer. Unfortunately, computerized counting opens up great opportunities for count-fraud, which is one reason why I don’t propose it or any other rank method for official public political elections.
- What if we counted the rank-balloting election by hand anyway, even though it would be a tremendous job? The problem is that the tremendous amount of count-labor would still provide the opportunity for successful count-fraud, as compared to the much simpler and easier Approval count.
- In that regard, IRV is even worse than many of the other rank methods including, a personal favorite, Improved Condorcet Top (ICT). You see, IRV is not “precinct-summable”; in a rank-count like ICT, the various precincts can count some relevant sums, and send those sums to a central counting location where those sums could very quickly and easily, with minimal computation, be counted to determine the winner.
That isn’t true of IRV. IRV isn’t precinct-summable. In IRV, it’s necessary for the central count to constantly refer to the rankings at each stage of the count. This would be a count-fraud-security nightmare and it would be a field-day for count-fraud. IRV, thus, has a much greater opportunity for successfully undetected count-fraud.
- Favorite burial incentive: Many rank methods provide incentive for voters to engage in “favorite-burial”–voting someone over their favorite. As I’ve already discussed, Plurality, our “Vote-for-1” method that we currently use in our official elections, is a prime example of a method that provides an incentive for favorite-burial voting.
Ask someone who votes Democrat, if that Democrat is really honest, trustworthy and their genuine favorite. They’ll probably tell you that you must pragmatically vote for the Democrat, and abandon your genuine favorite in order to help the Democrat beat the Republican.
Not all methods are like that. With an Approval voting methodology and with ICT there is no such problem. No one ever has any incentive or need to vote someone over their favorite. No one ever has any reason to not give to their favorite(s) the top rating, or the top ranking.
Instant Runoff Voting is Poor Choice
If you haven’t yet voted in Democracy Chronicles’ two presidential polls using Approval and ICT then I invite you to do so. You can find the Approval style Presidential poll in the left column on the front page of Democracy Chronicles but be sure to vote before the poll closes on August 22. You can find the ICT Presidential poll by following the link here.
Instant Runoff Voting imposes a strategic need to bury one’s favorite, and does so considerably more frequently and flagrantly than many other rank methods such as the ordinary Condorcet voting system. What would be the point of working for, and getting, a new voting system, if that system would retain the unnecessary problem possessed by our current Plurality system? Count-fraud and favorite-burial? No thanks.
David Cary says
Ossipoff tries to fear monger about Instant Runoff Voting. But his arguments fall flat.
Some implementations of IRV have set new standards of transparency for elections. In some cases, that transparency and an ability for members of the public to independently verify the tally has resulted in correcting errors in plurality contests!
Plurality voting is about as simple to count as any voting method, but even in the popular U.S. Presidential elections this November, which will all use plurality voting, nearly all of the votes will be counted with computers. Computers do add extra dimensions to the problems of election fraud, but they also can contribute to the solutions. In any case, Ossipoff greatly exaggerates any additional exposure that use of IRV necessarily contributes and he incorrectly claims that IRV has to be counted centrally.
All voting methods are subject to strategic voting. Ossipoff offers no explanation for why one technique (favorite burial) is necessarily worse than others, or why the type of technique is necessarily relevant. As a practical matter, opportunities for strategic voting are less common with IRV than they likely would be for Condorcet methods or approval voting. This is one of the reasons that Condorcet methods and approval voting are not used for public elections. It is also one of the reasons that voting with IRV is often easier.
Rich says
An more in-depth, unbiased analysis on Instant Runoff Voting can be found at the Elections Mathematics website:
https://electionmathematics.org/ucvAnalysis/US/RCV-IRV/InstantRunoffVotingFlaws.pdf
David Cary says
That paper is not unbiased. It is plagued by factual and logical errors. It offers a very distorted comparison to the alternatives discussed. The perspectives on election auditing are dated.
Michael Ossipoff says
David Cary says:
Ossipoff tries to fear monger about Instant Runoff Voting. But his arguments fall flat.
[endquote]
That remains to be seen, based on the validity of Cary’s justification for that statement.
Cary says:
Some implementations of IRV have set new standards of transparency for elections.
[endquote]
Nonsense. A more elaborate count is much less transparent than a simple count.
Cary says:
In some cases, that transparency and an ability for members of the public to independently verify the tally has resulted in correcting errors in plurality contests!
[endquote]
Well, I don’t know how they count Plurality in whatever places Cary is referring to, or why they had made errors in their Plurality count.
Cary says:
Plurality voting is about as simple to count as any voting method
[endquote]
Again, nonsense. Perhaps Cary didn’t notice the flowchart illustration in my article. Compare that to this:
In an Approval election, for each candidates, count how many ballots mark that candidate. Elect the candidate who is marked on the most ballots.
Car says:
, but even in the popular U.S. Presidential elections this November, which will all use plurality voting, nearly all of the votes will be counted with computers. Computers do add extra dimensions to the problems of election fraud, but they also can contribute to the solutions.
[endquote]
Another uselessly-vague statement. How does computerized count contribute to solutions, while providing enhanced opportunity for undetected count-fraud? …while making the count publicly-unobservable?
And, even if a computerized count is used, a simpler computer program is easier to check on than a more elaborate one. A program that counts Approval is a program that does the simplests task: It merely increments a total when encountering a ballot that marks a candidate.
With most voting systems, a computerized count is optional. Plurality and Approval, for instance, could be handcounted. In fact, Condorcet, being precinct-summable, can could be handcounted too, though they’d require considerably more count-labor than would Approval. The pairwise vote-totals could be counted at the precinct-level. It would be a relatively labor-intensive handcount, but it could be done. These pairwise vote-totals could be sent to the central count location, where each candidate-pair’s pairwise vote-totals would be summed, and the winner determined.
That won’t work for IRV. As I said before, in my article, IRV has to look at and process the ballots at each stage of the count. In a presidential election, the counting must be national. 100 million ballots, each examined and processed at each stage of the count. No, that isn’t a handcount.
If anything, then, the ridiculous count-fraud field-day that IRV would bring is even worse than I implied in my article.
For state offices and district representatives, the size of that central count isn’t as great, but the same problem remains, even if on a somewhat smaller scale.
IRV is much worse than any Condorcet version, inclusing ICT, in that regard. And incomparably worse than the simple Approval count.
Cary says:
In any case, Ossipoff greatly exaggerates any additional exposure that use of IRV necessarily contributes
[endquote]
Cary forgot to say which of my statements, in particular, was an exaggeration, and why.
Rank-counting is inherently much more elaborate than an Approval count. And IRV’s non-precinct-summable count is much worse than a Condorcet count, where, at least, it’s simple summation, mostly locally do-able, even if there’s more of it than in Approval.
I stand by my statement that IRV would be a count-fraud-security nightmare. Many agree. If Cary wants to challenge a statement that I made, then he needs to be a lot more specific about what he means.
Cary says:
and he incorrectly claims that IRV has to be counted centrally.
[endquote]
Ok, Dave, tell us why you think that IRV is precinct-summable, and how you’d facilitate the IRV count, or conduct it locally at precinct level instead of centrally. You seem to rely on vagueness. You need to give specifics.
Cary says:
All voting methods are subject to strategic voting.
[endquote]
That’s common knowledge in voting system discussion. But methods differ drastically and dramatically, in how much of a strategy problem they cause.
Cary says:
Ossipoff offers no explanation for why one technique (favorite burial) is necessarily worse than others, or why the type of technique is necessarily relevant.
[endquote]
On the contrary, I devoted considerable space to that, in my Approval article at Democracy Chronicles. It goes without saying that when voters are strategically-compelled to bury their favorite, that makes nonsense of the election and its result.
If Cary wants to claim that some other strategy problem, incentive or need is more imporant, then, again, he needs to be specific. And if Cary wants to say that IRV has some property that’s so valuable or important as to justify IRV’s favorite-burial need, then he needs to be specific about that too.
Cary says:
As a practical matter, opportunities for strategic voting are less common with IRV than they likely would be for Condorcet methods or approval voting.
[endquote]
The problem isn’t strategic “opportunities”. The problem is strategic compulsion or need. There is no strategic need more drastic than the need to bury one’s favorite. And nothing else distorts the public’s indication of what they want as much as the ridiculous and avoidable favorite-burial need of IRV. Approval has no such need.
(Neither does ICT, but I don’t recommend any rank method for actual official public political elections, for the reasons that I’ve given)
Dave says:
This is one of the reasons that Condorcet methods and approval voting are not used for public elections. It is also one of the reasons that voting with IRV is often easier.
[endquote]
IRV has been very busily promoted around the country by an apparently well-funded organization. Approval has no such organization promoting it. There’s much resistance to any new voting system, and enacting one takes a lot of work. IRV has had a national organization to do that work. I commend the hard work of the IRVists, but their work has been misguided.
Michael Ossipoff
Anonymous says
Seems like Cary’s statements don’t hold up to scrutiny. a very good point by point.
Adrian Tawfik says
Thanks for giving us your thoughts. Be sure to check us out at Facebook and Twitter so you can tell all your friends!