Thursday, September 4, 2008

Raleigh, NC overwhelmingly endorses Approval voting!

Met up with some friends last night at the Flying Saucer. It was "glass night" and you could vote for a presidential candidate (Obama or McCain) by buying their glasses. You could buy more than one, so I guess this was "approval voting" night at the Saucer - no IRV for these people.

The voting commenced at 7PM, and Obama had the lead until we left, varying between 58% to 60%. A crew of Obama volunteers came in. A McCain team came in, consisting of a rather constipated-looking young man and two Cindy McCain clones. We were surrounded by McCain supporters - some of whom couldn't believe that two small business owners (Jesse and myself) could possibly be Obama supporters.

Some Libertarians came in and were pissed there were no LIB glasses, or write-ins. You could recognize them because one looked like he lived in a cabin in the woods for years, and the other wore a "Munger for Gov" t-shirt.

When we left at 8:50, it was 223 Obama to 142 McCain. At closing time last night (verified by me by phone), and the result was 251 Obama to 187 McCain - Obama lead through the night and defeated McCain by 57%-43%

I am taking that as a good sign - and an endorsement of approval voting that produced an actual majority winner with no one arguing about it! No one needed a calculator to figure out who won, there was no complicated sorting of glasses, where some of them might break (so they wouldn't get counted), and no calculator error. No one came in and faked an accent or asked leading questions to get you to buy a glass or two or more.

This was a ringing endorsement of Approval Voting over traditional first past the post and IRV since no one objected to being able to cast as many votes as they wanted to. Money was not a factor in the Obama victory - no deep-pocket Republican came in and bought enough glasses for themselves or for other voters to keep up with Obama - there was just not enough interest in McCain to justify that sort of election trickery.

And even though there were only two glasses, there was plenty of choice to go around - you bought the glass, but could put any one of 200+ beers in the glass. So there were no crybabies bitching about not having choice or how people couldn't vote their hopes and dreams. Everybody supported the candidate and drank the beer of their choice!

Tuesday, September 2, 2008

Interesting pro-IRV site that talks about problems with centrallized processing of IRV ballots

Did more searching of google using "IRV sucks" as my search criteria, and this popped up - for general site, and this particular link for hand counting IRV:

5.5. How could you do a hand count for IRV?
If you have all the ballots in one place, it's easy. Separate the ballots into counted piles according to the first choice marked. If no one pile is 50%, take the smallest pile, cross out the first choice on those ballots, and redistribute. If any have no choices left, put them in an "exhausted" pile. Continue the process until one candidate's pile is over 50% of the ballots NOT in the "exhausted" pile.

The problem is that you have to have all the ballots in one place. You can transmit them physically or electronically, but you need every individual ballot. However you transmit them, it's an opportunity for fraud or incompetence.

If you don't have the ballots in one place, you can do the first count, transmit the results to a central location, wait until everyone else finishes and the results are all in, recieve the word of who to eliminate, recount those ballots, retransmit the results, wait again... It's a lot of chatter back and forth, and a lot of waiting for the slowest counter, not just once, but several times.

This is one of the biggest problems that I see with IRV in North Carolina or anywhere else - the fact that vote counting will now rely on central processing, not just central reporting. This seems to make sense for paper ballots, but here's where it gets much more difficult if not do you do the manual counts, recounts and audits of IRV elections?

5.11. So using the same mathematical language in Arrow's theorem, what are the flaws in IRV? In VOTE-123?

Both of them have problems with independence of irrelevant alternatives (IIAC). This means that they can be manipulated by controlling the pool of candidates. It also means that the voters can change the results by strategically ignoring (down-ranking) certain candidates who have no chance.

5.12. Wait a minute. Up above you say that under VOTE-123 "You can always vote honestly, period." But here you seem to be contradicting that by saying that no voting system can be perfect that way.

OK, you caught me. There's a theorem related to Arrow's theorem (the 'Gibbard-Satterthwaite theorem') that shows that there is no voting system without some strategy - that is, some situation where an omniscient voter could vote dishonestly and thus increase their satisfaction. But just as you can mathematically prove that there's always a situation like that, you can mathematically prove that (especially given the right tiebreaker) under VOTE-123 it is always too risky to try it. Specifically:

Strategic voting only works under VOTE-123 if, either before or after the strategic changing of votes, there is a "Condorcet tie" - what I've been calling a three-way tie. This situation is very rare - as I said, it involves nearly-balanced voting blocs which not only disagree about which candidate is better, but also about which candidates are similar to each other. Most especially, it is very, very hard to predict accurately using polls. Most people will never make the effort to reliably tell pollsters who their second choice will be, and in such a closely-balanced situation the tiniest polling error could throw off the poll's results. And if you switch around your vote based on a poll that's wrong, you're more likely to end up hurting than helping yourself - especially since your plans will probably leak out and people can plan a counterstrategy.

So, you can still say it categorically: under VOTE-123, there is never a good reason to vote strategically.

And from the same source - they claim that FairVoteSF coughed up $87,000 to push the pro-IRV Proposition A in 2002.

2.5.2. San Francisco, 2002: IRV vs. Lobbyists
For a more modern case, I plan to do an analysis of the San Francisco vote on Prop A (to use IRV in city elections) which passed in 2001. There were two committees against Prop A: one small committee of honest skeptics, and one that was run by Jim Sutton, the guy who's the lawyer for all the big-money campaigns in San Francisco. He used his whole bag of shady tricks to run a last-minute lying smear campaign without having to disclose anything about his funding until long after the election is over. (Once his funding disclosures come out in August 2002 and I have time to analyze them, I'll give you the details.)

Preliminary numbers from the Department of Ethics Forms 460: "No on A", a small committee formed by some people honestly suspicious of that the SF department of elections couldn't handle IRV, had a total budget of about $500.

It's easy to understand why some people felt that San Francisco Dept. of Elections couldn't handle IRV. One posting on here from the FairVote site shows they had doubts that SF could handle IRV in 2004! And it really hasn't gotten any better - between the reports that San Francisco still doesn't have a certified method to tabulate IRV votes, the civil grand jury that states there are still voter education problems after 4 IRV elections, as well as the month's-long IRV tabulation - they still have much to work out after 4 IRV elections.

Do we really want to buy into all that hassle? We must investigate cheaper and less complicated alternatives to IRV to deal with the occasional need for runoff elections. San Francisco has proved IRV isn't the way for NC to do it.

"San Franciscans for Voter Rights", an anti-IRV PAC formed by a big-time downtown lobbyist, declared in the last 16 days before the election to try to get around reporting requirements, sent out lying hit pieces using about $77,000. This originally came entirely from downtown businesses, most of it came funnelled through big PACs (such as $61,000 from the "committee on JOBS") which have a long term reputation for backing corporate-written bills (for instance, insurance industry bills, etc.).

If you add up the other contributions of those same large PACs to various "voter guide" organizations (the "so-and-so Democratic club", etc.) which all conveniently had a "No on A" stance, there are over $100,000 EXTRA (in addition to the amounts above). There were obviously other issues on the ballot, but given the way many of these "voter guides" highlighted their measure A "NO" with bold face, I think that it's fair to count at least 10% of this as anti-A money; so count this as another $10,000

Add in 1/2 of the $32,000 staff expenses, January-March, for "San Francisco Chamber of Commerce 21st Century Fund", which although it is supposed to be a "general" lobbying organization took up no other issues in this race: another $16,000

Total anti-A money: over $100,000

For balance, the pro-a money:

"Fair Vote SF", a pro-A group, got $87,000 from small contributions from individuals, most (about $55,000) of it funneled through national or state PACs that see voting reform as important (primarily Center for Voting and Democracy).

Using the same analysis as above (what else did the pro-A PACs do with their money?) there is no extra money to report. Obviously there were pro-A "voter guides", but these voter guides did not get money from clearly pro-A PAC's. Not surprising: generally, the forces of good have simpler accounting practices than the forces of evil. (For full disclosure - the CVD did give $2,713 to "Matt Gonzalez for Supervisor")

Total pro-A money: $87,000

Pro IRV sources in FL claim that FairVote contributed thousands if not tens of thousands in money and staff time to push IRV in Sarasota. I wonder how much they spent to push IRV in general and in Cary and Hendersonville in particular that never got included in the cost of implementing IRV?

Monday, September 1, 2008

They knew in Vertmont back in 2006 that IRV wasn't all it was cracked up to be!

After all the depressing stuff on Gustav today, I decided to have some fun. I googled "IRV sucks" and came up with some fascinating stuff. One of them was this entry in the Vermont Daily Briefing from May 2, 2006 -

Here is an excerpt:

During Burlington’s last mayoral election — the city’s first use of Instant Runoff Voting — Political Science professor Tony Gierzynski mobilized an army of exit-pollers and gathered massive amounts of data.

This data he then crunched. Crunched it real good.

It took a great deal of time and foresight, but the results were utterly unique: no one had done such a study anywhere in America.

This is true - they sure as hell haven't done anything like that in NC. And I don't see anything in the study where they tolerate a paid IRV advocate doing the survey work, both deviating from their instructions and faking a southern accent.

So I went to the survey - located at and, as usual, I went to the end where the juicy stuff is. Here is some of it:

The relationship between education levels and awareness and understanding of the IRV ballot in our exit poll is similar to the findings of the exit poll conducted during San Francisco’s recent experience with IRV (see discussion above) and is one of the main concerns with this method of voting. As the experience of Florida in the 2000 presidential election demonstrated, certain voters are likely to have enough difficulty with complex ballots so that their votes do not end up counting. The percentages of people who were unaware of IRV or found the ballot confusing in the Burlington election were low even for the lowest levels of education (undoubtedly due to the City of Burlington’s effort to educate voters on IRV). The number of confused voters represented by those percentages, however, would be much greater in elections in larger cities or in statewide contests. Additionally, the higher level of voter turnout in statewide elections means that a larger proportion of the electorate would be composed of groups that, according to both our results and those of the San Francisco exit poll, had more difficulty with IRV, namely, those with lower levels of education. To illustrate, the percent of eligible voters casting ballots for governor in Vermont in 2004 was 65.1%25 compared to the 30.3% turnout in the 2006 mayoral election in Burlington. Only 12% of voters in Burlington’s mayoral election had a high school degree or less, while 26% of voters in the 2004 presidential election had a high school degree or less.26 Because they represent the group that had the most difficulty with IRV, a higher percentage of voters with a high school degree or less would undoubtedly inflate the percentage, as well as the number of those uninformed about IRV and/or confused by it. In other words, there is a good possibility that the difference among voters based on education levels would be intensified in an election with a higher voter turnout.

Here is a place where IRV advocates point to success in IRV elections, yet a study done in their own version of the Institute of Government shows there is a good possibility that IRV could cause voters with less education to become confused by the IRV ballot.

So what did that report conclude?

The exit poll results do, however, raise a couple of concerns about IRV. There appear to be both education and partisan differences in the reaction to IRV. The relative lack of awareness and confusion voiced by those with lower levels of education suggests that IRV has the potential to engender some inequities in the electoral process based on class. The partisan divide found on IRV in Burlington — Progressives and Democrats liking the system, Republicans disliking it — poses a problem for the perceived fairness of elections and the legitimacy of those elected.

While a sound argument can be made that IRV functions in a manner to select candidates based on majority preferences, the minority party may see it as an unfair changing of the rules of the game that deprives them of a chance of winning when their opposition is in such disarray as to offer multiple candidates. Were any of the potential voting paradoxes discussed above to arise in an election (especially if a Republican candidate had the lead in the 1st round and lost after the second round) there would surely be attacks on the legitimacy of that election in the press. Then there will be a real test of the public's understanding of IRV.

Other questions remain to be answered regarding the effect of IRV elections. It is unknown how IRV affects campaigning. It may ensure more congenial elections since candidates would not want to alienate any potential 2nd place votes from supporters of their opponents. But, it could also make it so candidates play down their policy differences for the very same reason—not wanting to alienate any potential 2nd place votes—making it less clear to the voters what their choice really means in terms of governance. Nasty campaigning, it should also be noted, could simply go underground as it may have in the Burlington mayoral contest. It also remains to be seen how IRV affects voters’ decisions. Despite claims of its supporters to the contrary, IRV does allow for, and even encourages strategic voting (as opposed to pure preference-based choices). What voters’ strategy would be and how it might differ from their calculus in the typical single-vote system used in the US is unknown, as is the way that such differences might affect the outcome of elections.

In the end, elections are about building governing majorities. IRV offers an opportunity to have an electoral majority without doing the hard work and compromise necessary to build it. How that dynamic might affect the ability of elected officials to govern is a whole other matter to be seen.

Recent readings of mine suggest that IRV could exacerbate the problems we are having getting a census government - it could lead to a "balkanization" of American politics and make matters worse, not better. Why should we experiment with a voting method that might make matters worse, not better.

That is why this voting activist doesn't favor experimenting with votes using IRV to see if it can be made to work. The laudable goals of IRV could be achieved without all the risks that even the NC State Board of Elections knew were posed by IRV way back in March 2007 - even before the first communities were considering taking part in the first round of the pilot program.