Reprinted from the AP Wire
By Paul Nowell
Using computer software that models how consumers choose everything from fast food restaurants to mortgages, a trio of economists is tackling a daunting task: forecasting the college choices of the nation's top 100 high school football recruits.
"They say raising teens is like nailing Jell-O to a tree," said Allen Lynch, one of three researchers who created the computer model that spit out predictions in anticipation of Wednesday's national signing day. "Predicting the behavior of teens is always a difficult task."
While the signing day model is just for fun - and a little free publicity for its developer, SAS Institute Inc. - there's a serious payoff for marketers and other business people.
After all, if a computer program can predict where 100 teenagers will decide to go to college, perhaps it can also predict what new movie will have their classmates packing the multiplexes on Friday night or what new clothing trend might catch on in the nation's high schools.
Recent applications of the underlying technology have included helping draw floor plans for a designer shoe retailer, determining what combination of treatments can best keep diabetics from landing in intensive care and testing airplane reliability.
"A real world example would be if a bank wanted to understand why a customer would want to choose their credit card," said Anne Milley, manager of analytical strategy for SAS, based in the Raleigh suburb of Cary. "Is it the brand name or the interest rate, or is it because it doesn't have a fee?"
The answers to these questions can help bank executives decide what to stress in a marketing campaign, she said.
Lynch, an economics professor at Mercer University in Macon, Ga., developed the model along with Mike DuMond, an economist for the ERS Group in Tallahassee, Fla., and Jennifer Platania, an economics professor at Elon University in Elon.
Like a similar SAS program that predicts which teams will make the field for the NCAA men's basketball tournament, the signing day prediction has broad popular appeal.
"It's a nice way to package a very sophisticated modeling technique in a palatable format," said Lynch.
Platania said decisionmaking models are crucial for companies seeking an edge in the competitive marketplace. "Even if you can increase your odds of success by 1 percent, it can bring in millions of dollars in sales," she said.
Lynch was also the mind behind SAS's "Dance Card" model, which projects the teams that fill the at-large berths in each year's NCAA tournament.
Since 1994, "Dance Card" has correctly predicted who will receive 352 of the 376 available at-large berths. That's a success rate of 93.6 percent.
National signing day offers a tougher challenge, Lynch said. In the football recruiting model, the high school stars are likely choosing from among multiple scholarship offers.
Using data from Rivals.com, a series of Web sites dedicated to college football, the economists write a predictive formula that ranks factors like a school's national ranking and whether the team plays in a major conference. Things that can't be measured for use in the formula are less obvious factors - like whether a sibling or high school teammate went to the school in question, Lynch said.
"We look at things like if the player wants to be close to home and if they want to play a lot," said Platania. "It could be that his parents don't want him to go to a school, so he signs there."
Between 2002 and 2004, the model accurately predicted where a student-athlete will decide to go to college about 61 percent of the time.
Through Monday, Lynch said, this year's program was doing better - it had correctly predicted 58 of the 78 verbal commitments made by the top 100 players, or 74 percent.
"While it may sound to some people that this is not much better than 50-50, you have to remember that these recruits are usually picking from four schools," said DuMond. "It's like throwing darts and you hope to do better than hitting the mark 25 percent of the time."
Lynch cited prep star Fred Rouse of Tallahassee, Fla., as an example. Rouse's hometown program, Florida State, is very interested in signing him - but so are several other top programs.
"If he were not one of the great players, Florida State would have an inroad," Lynch said. "But he can pick if he wants to go there or take the offer from Southern Cal or someone else, so that makes our job a little tougher."
The latest prediction from Lynch and his team had it a toss-up between Rouse becoming a Seminole or signing with Alabama. After weighing offers from those schools along with Texas and LSU, Rouse signed with Wednesday with Florida State.
As maddening as it can be to try to determine the thinking of 100 teenagers, it's also a labor of love for Lynch and his colleagues.
"All three of us are huge college football enthusiasts," he said.