Winning American Idol: try to be last

If you are appearing on American Idol or the X-Factor, try to be one of the last to sing. That's the conclusion from a new paper presented at a University of Westminster seminar today. Lionel Page and Katie Page look at an important topic - the evaluation of a sequential order of performances. Their paper, Biases in sequential performance evaluation, a field study on the Idol series (PDF), drew on a large dataset based on the ranking of contestants in live pop Idol shows in 8 countries (Australia, Brazil, Canada, Germany, India, Netherlands, UK, USA). That's 1,522 performances from over 165 shows. They found two main biases which influenced overall sequential performance ratings:

Our results suggest that the two mechanisms, memory and direct comparison, both play a role in the order bias. With respect to memory it appears that both primacy and recency effects are implicated when sequentially evaluating performance. Irrespective of ability, contestants who perform first are more likely to be positively evaluated than those who come in second and third positions, which provides evidence of a primacy effect. Contestants who perform in the later serial positions (particularly last position) have the largest advantage with respect to positive evaluations, implying a strong recency effect.p>

...The second bias we demonstrate is a direct comparison effect with the previous contestant. Specifically, one's performance evaluation is influenced by the evaluation of the previous contestant. If you perform after a weak contestant there is a bias such that you are more likely to be evaluated poorly than if you perform after a strong contestant. Therefore, we find evidence for an assimilation effect with respect to sequential judgments.

...Overall, we show that these two effects both operate and are important explanatory mechanisms in the evaluation of sequential performance.

Something to ponder when you go speed dating, or get a call to say you've been shortlisted for a job interview and are asked what slot you'd like.

UPDATE:

WSJ's Matt Phillips has obviously read my post, writing about the paper at the Journal's real Time Economics blog: Last Shall Be First in Idol Economics

Comments

[amselly]

Maybe they schedule more popular contestants (and more likely to win) later in the show to keep people watching until the end.

[Jay]

Ground breaking! I think the second effect was discovered a long time ago by predatory male college students looking for a female that was recently hurt emotionally by a significant other.
If the conversation starts along the lines of...
"You don't look too happy, can I help you?"
"I just found out my boyfriend was cheating on me!"

you really can't go wrong if you are the guy in this situation.

[Jay]

Whoops, misread the quoted text. Instead I have to greatly disagree with the second conclusion. I'd much rather have a job interview after the fool that couldn't explain what an R2 statistic is than after the one that did so well he all but landed the job.