Sunday, May 21, 2006

idol curiosity

Okay, so prompted by developments mentioned in the bottom of my last post, I devoted some more of my insomnia to American Idol research, despite having never watched the show. I was struck by this story:
Elliott Yamin, of Richmond, Va., received 33.06 percent of the vote. Hicks, of Hoover, Ala., and McPhee, of Los Angeles, got 33.26 percent and 33.68 percent of the votes, although it wasn't revealed which contestant had more.
Three candidates, and they all get vote totals within six tenths of a percentage of one another. Hmm. I mean, people's blogs made it sound like the judges and everyone else sensed that Yamin was going to finish third, but he loses only by 1 in 500 votes?

I looked up the final vote totals from the first four seasons of American Idol:
Clarkson def. Guarini 58-42
Studdard def. Aiken 50-50
Barrino def. DeGarmo 51-49
Underwood def. Bice 50-50
For democratic elections with two parties, there are rational choice explanations you can invoke for why elections should tend to be very close. (Basically, a party that is winning by too large of a margin should have pressure to become more ideologically extreme and thus make the next election closer.) I see no straightforward application of this logic to American Idol. So then: why are these elections repeatedly so exceptionally--or, to use a different word, unnaturally--close? What I am saying is that these numbers have been repeatedly so close as to demand some kind of explanation, and so what is it?

I'm serious here. I can think of various theories, but I'm interested if anyone else has ideas. For certain of my ideas, interesting additional evidence would be provided by the vote totals for Idol in other countries (it still runs in other countries, right?), and seeing if those also evince a tendency to be extremely close elections.


Anonymous said...

My theory: AI has hired Diebold to count the votes for them, and for the same reason that Ohio wants Diebold voting machines -- to skew the vote.

islander said...

I get the sense that true fans call in votes continuously during the entire 2 hours that the lines are open. Since it is hard to get through in the first hour of voting, I'd suspect that the difference is the number of folks willing to spend 2 hours making continuous calls.

A+ said...

It's rigged. I'm not trying to be cute or glib here, I just have a feeling it's rigged.

Bill said...

I think the pressure of the median voter stuff is still going on in this case. The thing is is that there is a selection bias, the contestants are not sort of randomly drawn, but instead are selected largly on their ability to appeal to a wide audience. The Downsian theories of voting supply that parties will move to the middle to capture the vote. Replace parties with Simon Cowell--viola--economic democracy in action.

Bill said...

two edits. First, viola should be voila. Second, largly should be largely. I may have missed others. Ignore those.

Anonymous said...

It is nearly impossible these days to register a vote on American Idol. (Trust me. I try all the time; the lines are nearly always busy.) I have it on somewhat good authority that when this situation occurs, the number of votes cast for a particular option a telephone poll is highly directly correlated to the number of lines available to collect calls/votes and the length of time the lines are open and has a very low degree of correlation to the actual popularity of the options.

One can only assume that all candidates have the same number of lines dedicated to receiving calls and that all the lines are open for the same amount of time.

As a result, the total number of votes will come in almost identical, and the result is little better than a crap shoot.

Anonymous said...

Oh yeah, I forgot to mention - there's a strong suspicion that it's rigged too.

But I have no reliable source to confirm this, so the capacity argument prevails for the time being.

Bill said...

To anon 1:32,

That assumes that the problems getting votes for candidates will be biased against the more popular option. Problems getting votes in to candidates should affect all callers equally, and the end result should not be that different than if it was entirely costless to enter in a vote.

Anonymous said...

Dear Bill,



Anon 1:32

Bill said...


From what you said, it seem random who actually gets in to vote. If that is actually the case, it would not change anything in the final vote totals. As long as it's equally hard for both sides to enter their votes, than the results would be exactly the same then if it wasn't the case that it was challenging to actually get a vote in.

Hopefully that clears it up.

jeremy said...


This is not correct. AI has different lines for different contestants. If those lines are being used to capacity, then it's "random" which caller voting for X gets to vote, but not whether someone voting for X or Y gets to vote. Instead, it would tend to equalize the numbers for the two candidates. An implication of this would be that, in an election where lines were overcapacity for the whole period, the election would go to the candidate whose voters spent the least time between having their vote registered and hanging up.

Bill said...

Sorry Anon, I missed the tag where you added the assumption that there must be dedicated lines. My missread.

A question though: what about the votes sent in by texting on your phone?

While USA today may not be the most accurate source on the face of the earth, AT&T claims that they submitted the 2.5 million votes that came in, and that none of them were rejected. If the phone lines for both candidates are stuffed to the gills with voters trying to get in, then the texted votes would provide the difference. But, I don't know how long this system has been in place though, so I can't speak to the history of the voting.

I still think the Economic Theory of Democracy stuff works here (I never thought I'd be saying that when I read it, damn). Especially with the way that individuals are knocked off in consecutive rounds. What you are left with is two candidates who appeal strongly to the middle. Considering that the initial selection sequence involves screening them for their ability to sing, and their broader appeal to an audience, the outcomes are not too suprising.

Anon 1:32 said...

I don't doubt that there's a "race to the middle" in the contest. The whole point of the show is to find someone pleasingly MOR who can sell assloads of records. Anyone who is at all distinctive or out of the ordinary is quickly eliminated (remember Amy Adams? Jon Peter Lewis?).

My point is that inspite of the median voter hoobeejoo, the general voter preferences are not as close as the actual voting results imply. The closeness of the actual results is simply a technical glitch.

You also bring up a good point about the text messaging votes. However, the number of text messaging votes in relation to the total number of votes cast is so small as to dwarf their impact.


Assume the following in a two-candidate vote:
1) Telephone votes are equally split between the two candidates.
2) Text-message votes are 10% of the total number of votes cast.
3) Text-message votes split 60%/40% between the two candidates.

In this scenario, the final split of the total vote between the two candidates is 51%/49%.

If we assume that the text-message votes accurately reflect the preference of the population as a whole, then there's no problem. The preferred candidate wins (albeit by a drastically reduced majority).

However, the cost of casting a text message vote is not the same as the cost of casting a telephone vote. Telephone votes are made to toll-free lines, and the only significant cost to a voter is the time involved in dialing and waiting for acknowledgement of a cast vote. Text-message votes have basically the same time cost for the voter, but there is also an additional monetary cost associated with sending a text message - a nominal charge on your wireless bill, a reduction in the number of text messages available as part of a package, etc.

It's also important to note that only customers of Cingular/ATT Wireless can cast text-messaging votes.

So now we're creating two different sets of unrepresented voters: those who are not Cingular/ATT Wireless subscribers, and those who are but are unable (or unwilling) to pay to ensure their vote counts.

So the important question here is "Do we really want to leave the selection of an '*American* Idol' to a handful of Ma Bell devotees?"