Flemington, NJ, April 22, 2010 – A series of national studies conducted among 8,309 American Idol viewers revealed that the contestant who was “voted off” during the show’s elimination last night did not accurately reflect the viewers’ choice.
The Idol Democracy study was conducted during April 20-21, by HCD Research using its mediacurves.com website, to obtain American Idol viewers’ perceptions of which contestant should be eliminated on the April 21 results show. HCD Research is conducting a series of studies among a nationally representative sample of Americans, which parallel American Idol voting to determine the weekly winners based on a democratic, “one person, one vote” methodology.
During last night’s show, Tim Urban was eliminated from the competition. In contrast, viewers participating in the national Idol Democracy study, which employs a “one person, one vote” system, reported that Aaron Kelly should have been eliminated from the competition last night.
The American Idol voting methodology allows viewers to vote multiple times for multiple contestants. However, the Idol Democracy study only allows viewers to vote one time for one contestant. To view detailed results visit: www.mediacurves.com/americanidol/idoldemocracy
To register for weekly Idol Democracy studies go to: https://www.hcdsurveys.com/panel/join.cfm?source=idol2010
Among the findings:
Respondents were asked to pick the one contestant that they thought should advance to the next round of the competition:
|
Contestant |
% that voted for contestant to move on |
|
Crystal Bowersox |
44.5 |
|
Siobhan Magnus |
12.4 |
|
Lee Dewyze |
10.8 |
|
Casey James |
10.7 |
|
Michael Lynche |
9.6 |
|
Tim Urban |
7.3 |
|
Aaron Kelly |
4.8 |
Editors/Reporters: For more information on the study, or to speak with Glenn Kessler, president and CEO, HCD Research, please contact Vince McGourty, HCD Research, at (908) 483-9121 or (vince.mcgourty@hcdi.net). You can also receive updates from MediaCurves.com by following us on Twitter: http://twitter.com/mediacurves and Facebook: http://www.facebook.com/pages/Flemington-NJ/MediaCurves/86691908820
HCD Research is a marketing and communications research company headquartered in Flemington, NJ. The company's services include traditional and web-based research. For additional information on HCD Research, access the company’s web site at www.hcdi.net or call HCD Research at 908-788-9393. MediaCurves.com® (www.mediacurves.com) is a media measurement website that provides the media and general public with a venue to view Americans’ perceptions of popular and controversial media events and advertisements.
Week after week, you say that American Idol has gotten it wrong, but each week it's actually you who has it wrong.
You and American Idol are asking two very different questions. Your design only asks, "who should stay" - you don't have the data to predict who should leave.
With the multi-vote approach, it's not so much that people vote multiple times for the same person, it's that people vote multiple times for each of the contestants who they think should remain, and Idol gets the distribution of all who should stay down to the one who should leave.
If your "one person, one vote" question asked who should NOT continue, your results would be virtually identical to their results in terms of who should leave. Right now, statistically, the two "studies" should match at the top end of the spectrum but veer off at the lower end. However the top of the spectrum is so skewed, that asking the question that you do, who should remain, is virtually useless.
Put simply, American Idol's system solicits a vote for everyone who in the viewers opinion deserves to remain, therefore the obvious vote-getters get votes, the marginal contestants get fewer votes, but still valuable votes in the aggregate; and most importantly those with the least support leave - that's how their system was designed and for what they want to do, it's a good design. In effect, their data supports the entire spectrum from best to worst.
With your approach, if someone is conflicted between 2 or more contestants, they are forced to decide on only one; their possible secondary and tertiary votes, valuable pieces of information, making all of the difference with respect to lack of votes at the lower end, are lost. If you are only going to allow one vote per person, and you want to predict who should or will leave, you have to get a handle at the lower end data, hence my suggestion that you ask who should does NOT deserve to remain, or alternatively report what your data does show, who will stay - but obviously that would be boring.
By way of example, think of a panel of people asked to judge 10 paintings. Unbeknown to the panel, 5 of the painting were done by the top students from the nation's leading art institutes, and the other 5 were from a high school art classes. The panel is told to "vote for the best".
Do we have any information on which painting the panel thinks is the worst? No! It's certainly not the one with the least votes. The one with the least votes is the one painting that the fewest members of the panel think is the best. We have no information on second best, third best, and certainly not the worst.
If the fewest people think you're the best, does that make you the worst? Statistically, no. One week here and there it could turn out to be the same person but votes "for the best" at the lower end tend to be no better than random. In other words, some people will find each of the high school paintings "the best", but when voting for "the best" consensus on "the worst" cannot be inferred.