Marriage Poll (yes, again. Bite me.)

[crschmidt]

The Marriage Poll that I've been talking about for the past couple entries has changed significantly over the past few days from something that seemed (to me) to be seriously biased (95% against homosexual marraige) to a much more expected rate (about 54% in favor of allowing homosexual marriage.)

Around 10pm last night, they redid the site so that the results were only displayed on the page that you vote on, rather than on the other page. I have fairly accurate stats from 10pm Wednesday night until that time yesterday graphed. You can see from this that the results go through a nightly low in changing, and then increase during the day or afternoon.

http://peanut.sytes.net/marriagepoll/fri10pm.png

Note that this is a graph of percentage changes over 5 minute intervals. There are some intervals stored where the site is down, but I assume that during these times, no one is voting, so they're removed rather than left in as discontinuities. (Also, that makes formatting the data much easier for me).

In the process of following this poll, I have gotten over 1000 hits to the logs of 5 minute recent records. I have recieved emails, comments, and IMs thanking me, praising me, and in some cases, arguing with me.

3000 votes were removed from the poll at one point, which is why I started this. The AFA claimed these results were posted by an auto-responder, a fact which was later confirmed.

However, I still have doubts as to the validity of the results of this poll. I see trends which interest me and amaze me. For example, contrary to all other parts of the internet, conservatives on this topic appear to be online at 6am. From a comment of mine: http://www.livejournal.com/users/crschmidt/234947.html?thread=1028803

I have looked in every journal i could find linking to my site and seen how people are responding. Lots of people using it as an information source - it's an interesting demonstration of how a sub-meme within a meme can spread so much more easily. No one has to bring the topic up again - just add a link in a comment, and all of a sudden, I get 5-10 hits.

Also, the power of the internet, and LiveJournal specifically, is readily apparent in this poll. There are 4 times as many votes today as there were a few days ago, and I have seen information on this poll linked from so many places I can't even count them anymore.

It's incredible what the internet can do.

Comments

[sherm.livejournal.com]

I understand what you're trying to say/do here, but it doesn't change anything about the utter uselessness of this or any other online poll.

If I can vote 3,000,000 times, it's not a poll, it's a clicking contest.

While they're cheap and might provide general information about what certain groups of people are thinking, they're not statistically valid in any way and are (like this "association") total BS.

It's great that you can manipulate crap numbers to come out in favor of your point of view instead of a ridiculous one, but it doesn't prove the "power" of anything other than the ability of people to believe anything they see in print or on a screen.

[crschmidt.livejournal.com]

I don't think you can vote 3,000,000 times anymore - that was the whole point of a lot of changes they made. They put in a lot of support for unique identifiers, as well as carefully tracking who's voting, according to an email they sent.

Now, I haven't tried to fuck with the numbers, so I don't know. I'm all for trying to break things, but really, I'm not interested enough to try and screw up the results.

I don't think this poll is worth anything. If there's anything I've learned about online polls over the past couple days, I think it's that online polls are useless as anything other than a broadly sweeping demographic tool. I don't think this poll being sent to congress with highly biased results would have had any affect - nor do I think that sending it with unbiased results will do anything more. I dont' think it will change opinions. However, I do think it reflects opinions.

I feel that in general, online polls accurately represent the vies of the people who vote in them. I think that in this case, that means that the poll would normally represent the views of the AFA (and did, for the first small period of time that it was up). However, I think that the fact that results are much less biased now is an indicator of the fact that the people voting in the poll have changed: there is a larger percentage of voters who agree with the second option.

I don't think this will change anybody's mind. I don't think it will affect anyone's opinions. I don't think it's anything more than an interesting tool for finding out how a certain group of people are thinking.

But I do think that there are a lot of people who voted solely because this was spread around livejournal, and I think that this is an indication of the way that a meme or the internet can influence the group of people who are affected by something.

[rho]

What makes you think that it's less biased now than it was in the first place? It has one group, the AFA, who are highly biased. And then it has another group, those who were suitably morally outraged and wanted to be heard, and their friends. And this group is also, of course, biased in the opposite diection.

It's sort of like a poll to determine whether more money should be spent on sports or the arts, with people being asked outside an art gallery and a football stadium. It shows you the too extremes, but it in no way reflects the views of middle america.

[crschmidt.livejournal.com]

I think that the answer there would be obvious - averaging the two obviously makes it a less biased poll overall. Even if you're only taking two extremes, you're still going to get a LESS overall biased poll.

Now, do I think the poll actually indicates what some reasonable cross-section of people believes? Not at all. I think it's quite obvious that it indicates what the AFA believes and what LiveJournal believes. (And possibly some other sources, in both cases.) The results are obviously biased towards those two viewpoints. However, I feel that a bias towards two endpoints ends up being a lot closer than a bias towards one.

[rho]

That only holds if the relative proportions of the population as a whole are the same as the relative proportions of those taking the poll (iirc, that's called stratified sampling). You may (and probably do) end up with a result that is closer to the true result by having the dual-bias, but there's no way of being certain without actually doing the poll properly (for all I know, 95% of the US may be opposed to gay marriage, for instance). There's still enough bias that the results are entirely meaningless, in any real sense. All you're doing is changing from one set of meaningless results to another set.