Of all the social sciences, only economics knows how to cut to the core of a problem with rigorous methods and actually come up with real answers to some of society's most pressing questions. I really enjoyed Freakanomics , a fascinating look into how an upstart U. Chicago economist has been able to use economic tools to study everything from the organization of crack gangs to the kinds of names that people give to their babies. Why are black voting turnouts so low in some parts of the country? Why do women typically never marry down in terms of education? All of these questions are fair game for us.
It should come as no surprise, then, to see what a couple of friends of mine have decided to take a closer look at: whether certain characteristics make someone more likely to win American Idol. Forget global warming, this is the problem we need to be devoting our attention to. (Last year, this same friend and I created a formula for calculating a power ranking of the teams in our WAKA division based on kickball game and flip cup records, soooo... yeah, this is what we do).
Taking the results of all 5 seasons of American Idol, this crackpot team looked at the top 12 contestants for each season and categorized them according to sex (M or F), being overweight (yes or no), ethnicity, and home state. They then ran a regression of all these variables on the 60 contestants' ultimate finishes to find out if any of these characteristics contributed to how they performed. Here's what they found (most of the results are reprinted, with anyapologies to the author):
Table 1 - Analysis of American Idol Contestants
REG ProcedureModel: MODEL1
Dependent Variable: Rank
Number of Observations Used: 60
Analysis of Variance
Source/ DF /Squares /Square /F Value/ Pr > F
Model /4 /49.72031 /12.43008 /1.01/ 0.4091/
Root MSE 3.50393 R-Square 0.0686 Dependent Mean 6.51667 Adj R-Sq 0.0008Coeff Var 53.76871
Parameter Standard Variable Label DF Estimate Error t Value Pr > t
Intercept Intercept 1 7.20664 0.93790 7.68 <.0001 .
SexDummy SexDummy 1 0.01242 0.91053 0.01 0.9892 0.98835
EthnicityDummy EthnicityDummy 1 0.10870 0.90567 0.12 0.9049 0.99788
WeightDummy WeightDummy 1 1.06355 1.33985 0.79 0.4307 0.98640
SouthDummy SouthDummy 1 -1.72671 0.91053 -1.90 0.0632 0.98835
The conclusion? I'll spare you the technical details, but the number in red shows that all else being equal, someone from the South is more likely to finish 1.73 places ahead of someone not from the South. This is pretty consistent with the fact that all of the winners so far, and 3 of the 4 runner-ups, have been from Southern states. Tonight's finale? The Alabama boy vs. the LA gal... take it for what you will.
Now, there are a few problems with this work. First, we can definitely tell there's a correlation between hometown and performance, but it's hard to say if there's an actual cause and effect here; in other words, can we actually predict if a Southerner is more likely to win in the future, or are we just proving past results. Second, there really aren't enough observations to say with a lot of confidence that this means anything -- the model itself fits pretty badly. But the fact that they found anything statistically significant is pretty damn interesting. Of course, if you really wanted to go nuts, you could easily set up some interaction dummy variables, basically to see if, for instance, black Southerners (Ruben Studdard) or skinny females (Kelly Clarkson, at one point in time) have the advantage...
And Hilary thinks that people my age aren't productive enough during their workdays. Whatever.