Q&A About the election

[Jon Seattle]

Hi All,

Delia asked me to post to answer some questions people have about the election. I will also post some summary statistics later, but wanted to start off the discussion by answering a few questions. Please keep this technical, but I am glad to answer questions you have.

1. Does the number of candidates from a faction effect election outcomes?

No. You can think of this as two different (and not connected) elections.

a. Citizens rank factions and those ranks are used to allocate seats to each faction. Only the faction rankings matter in this phase of the election.

b. When you voted you also got to rank the candidates within your favored faction. This is an inside-the-faction election for faction leadership. It does not have any effect on the number of seats allocated to each faction, just who is favored to fill those seats.

2. What do the in-world machines do?

Those are more registation machines than voting machines. The SC provides me with a list of citizens who are eligible to vote. I use that list to initialize the database. When you touch the screen of an in-world voting machine, it contacts the server. The server then looks up the voter's name, and if your name is on the list generates a random password.

When you log in to the voting site, you must provide this password in order to vote. If the password matches the system will either provide you with a ballot (if you have not yet voted), or provide you with a record of your voting selection (if you have voted.)

3. Are votes counted as they are registered?

No. The online web-based program only collects votes and stores them in a database. The votes are delivered to the SC in a (tab delimited) data file that does not include the voter identity, and presents the votes in random order. With the file of voting data, I also provide a small program that does the seat allocation and candidate ranking according to the formula in the constitution.

The small program I provide can be re-run by members of the SC to check the results. The program is written in the Python programming language, and I send the complete source code so that they can examine it to make sure it works as the constitution requires.

4. What can someone do to understand Borda counts and the Sainte-Laguë seat allocation method?

Its pretty tricky, but I would start with two Wikipedia articles:

http://en.wikipedia.org/wiki/Borda_count
http://en.wikipedia.org/wiki/Sainte-Laguë_method

Sainte-Laguë is basically a system of proportional representation, but in our case instead of being proportional to the number of people, it is proportional to Borda counts. The combination has some complex interactions. Personally, I am very glad CDS is moving to the much better understood STV next election.

4. Can Jon release the voting data file?

Not on my own. This is up to the SC to decide. The tricky issue is that votes are confidential (the system is designed so that not even I see them.) It may be possible, even with non-identified voting data, under some circumstances, for someone to reconstruct how an individual voted. The SC needs to weigh the risks. If they decide to release the data I have no objections.

5. Who is Jon anyway?

Just a volunteer. I wrote the current voting system during my christmas holiday one year at the request of the RA, and run it under the SC supervision when they ask. I work as a Software Architect at Northwestern University in Evanston IL (real name is Jonathan Smith)

[Rose Springvale]

We've always seen the data for elections before, is there a reason we are not getting it this time?

Of course only SC people can post on the SC thread, so i'll just make this an official request to the SC to release the date to everyone.

Thanks
Rose

[Jon Seattle]

Hi All,

The SC has approved my posting the election data -- but the silly forum software won't let me attach the file, so I am linking to it here:

http://conchis.org/cds-election/votes.txt

The attached file is tab-delimited with one line per registered vote. The first three columns are the faction ids in preference order. The 5 through the end of the line are the candidate names in preference order. No voter ids are included, and the votes are in random order.

Best,

Jonathan

[Rose Springvale]

Thanks Jon. It looks like 57 voters by a quick count.

[Patroklus Murakami]

Thanks for posting this Jon. The election results are always interesting, these as much as any others. I note that the three factions which contested the election - CSDF, DPU and SEED - got roughly a third of the first preferences each. There should be no real surprise therefore regarding the division of the 11 seats up for election.

It's worth bearing in mind that all factions have roughly equal support in the community at this point as we work together this term.

[Rose Springvale]

More importantly, over half our citizens did not vote at all.

[Cindy Ecksol]

Thanks for posting this Jon. The election results are always interesting, these as much as any others. I note that the three factions which contested the election - CSDF, DPU and SEED - got roughly a third of the first preferences each. There should be no real surprise therefore regarding the division of the 11 seats up for election.

It's worth bearing in mind that all factions have roughly equal support in the community at this point as we work together this term.

Hmmm. I don't read it quite that way. A more precise and less misleading statement might be: "Among the less than 50% of our citizens who voted in the last election, approximately equal numbers had a first preference for each faction that contested the election." I supposed you could also say something a little less formal like "Among citizens in our community who were engaged and motivated enough to vote, approximately equal numbers had a first preference for each faction."

Cindy

[Patroklus Murakami]

Yup, those conclusions can be drawn too, but so can mine In the absence of any evidence to the contrary, it's fair to assume that the distribution of support for the three factions which contested the election in the population is similar to that of the citizens who turned out to vote. I can't readily think of a reason why that distribution would be different.

50% turnout is not that great. I think it may be a new low though I'd have to check the figures. My impression (based on my faulty memory having participated in about eight of these elections now) is that our turnout is declining over time from the seventies, then sixties to where we are now.

I think part of the reason was the electoral system, a factor that won't feature in our next election. The combination of ranking parties and then distributing them by proportional representation is 'rough sandpaper'. It hugely favours minority parties. Add to that the fact that factions had to have over 10% of the population behind them to participate and the outcome is virtually guaranteed. There really wasn't much incentive to work hard at getting out the vote under our old electoral system. Each additional vote (beyond the core which you hope will vote anyway) does not really get you that closer to an additional seat.

All of this changes when we move to Single Transferable Vote (STV) in the next election! It will be really important to get people to vote for you as an individual (even if standing as part of a faction slate of candidates) or at least rank you highly. I predict more campaigning and a higher turnout next time around. (Feel free to bash me with this if it turns out to be too optimistic!)

[Cindy Ecksol]

Yup, those conclusions can be drawn too, but so can mine In the absence of any evidence to the contrary, it's fair to assume that the distribution of support for the three factions which contested the election in the population is similar to that of the citizens who turned out to vote. I can't readily think of a reason why that distribution would be different.

Actually my conclusions are completely supported by the data. Yours are supported only by assumption (as you correctly note here). Your mindset may not allow you to envision reasons why the distribution would be different, but that does not validate your assumptions.

I can envision many scenarios that would result in your assumption being invalid. For instance, what if the vast majority of those who did not vote were citizens who had joined CDS less than 120 days prior to the election? Then the only valid statement you could make would be "Among those who have been citizens for more than 120 days there is an approximately equal first-choice preference for each of the three factions who contested the election." You cannot make any other statement because the preference among those new voters might turn out to be for one particular faction or another and there are enough of them to be statistically significant. What if citizens of one faction were the vast majority of those who did not vote, perhaps as some kind of protest movement? Then all you could say would be "Among citizens who are not members of Faction X there is an approximately equal preference for each of the three factions that contested the election." Those non-voters now BY DEFINITION strongly prefer a single faction and would skew the overall preference pattern. What if the vast majority of non-voters were those who do not believe their vote would make a difference? Then you could say "Among those citizens who believe that their vote makes a difference there is an approximately equal first-choice preference for each of the three factions who contested the election." And so on....the possibilities are endless, and some are not so trivial. Perhaps we should give them some thought.

Cindy

[Cindy Ecksol]

I just took some time to look at the election data in a bit more detail. Here's an interesting little chart that counts the number of first and second place rankings for each faction:

Code: Select all

            1st	2nd
CSDF	     21	 25      
DPU	      16	 25
SEED*       20     7

In the simplest counting fashion, CSDF received 46 total votes, DPU 41, and SEED* 27. I'd count that as a preference among those who voted for CSDF (40%, with DPU 36%, and SEED* 24%), not as "approximately equal preferences" for all three factions. Even if you give two points for a first place vote and one for a second place vote (as our system does - and 3rd place votes receive 0) then you get: CSDF 67, DPU 57, and SEED 47. That's an even strong preference for CSDF: the faction received 39% (DPU 33%, SEED* 27%) of the total points available. Not a majority, not a mandate, but a pretty strong showing even with the "leveling out" effect that our counting system imposes.

It sure would be interesting to understand more about those who did not vote. And I'm now VERY certain that we can't make any statements about what preferences our citizens might have in general, only about what preferences those who voted have demonstrated.

Cindy

[Patroklus Murakami]

Cindy

I suspect we are unlikely to agree

Your conclusions are entirely supported by data and mine rest on an assumption but my assumption is a reasonable one to make. The preferences of the community as a whole are (currently) not known. If someone were to poll everybody in the CDS we might get an answer. It could be that all the non-voters would have voted for one or other faction or none... We will probably never know. And, in any case, it's votes that count in a democracy.

The most reasonable assumption to make though is that the voter turnout is roughly representative of political opinion and preferences in the CDS. It's the simplest explanation for the result. Occam's razor would support my point of view!

That being said, one thing we agree on is that we should give some thought to why roughly half of our citizens did not vote. There will be some logistical/personal reasons - people were on holiday/too busy at work/otherwise busy in RL. There may also be some political reasons - Simplicity Party voters who felt the other factions did not represent their views. There may have been non-voters who were protesting (but why did we not hear from them and their protest?) There may have been people who thought their votes would not make a difference (and they would have been partly right, our old electoral system does not automatically yield proportional results).

Once again, I will predict a higher turnout next time. STV will force individual candidates to go out and get people to vote for them as individuals. No campaigning, no seats I expect that will take us back up to 60-70% turnout which I think would be a respectable figure.

[Jon Seattle]

Here is a first cut at it:

Jan 2010, elegible = 122, voted = 57 (47%)
Jul 2009, elegible = 71, voted = 39 (55%)
Jan 2009, elegible = 72, voted = 43 (59%)

So the trend has been down, but my guess is that this is mainly because of the percentage of newer citizens voting in the latest election. Personally, I am very glad we are going to STV.