Economics of recruiting

Publish date:

It's economics.

Using equations that contain more Greek letters than your favorite school's fraternity row, three economists have devised a way to predict the college choices of top football prospects. In two recruiting seasons (2005 and 2007), the College Football Recruiting Prediction model developed by Mike DuMond, Allen Lynch and Jennifer Platania has correctly predicted the college choice of the members of the Rivals 100 with 72.5 percent accuracy. The economists, who were published last week in the Journal Of Sports Economics, plan to predict the destinations of the nation's top 250 recruits on their Web site later this month, but DuMond ran the numbers of the three previously mentioned recruits to offer a taste of what to expect. According to the model, there is a 40.2 percent chance Pryor will choose Ohio State, compared to a 37.9 percent chance he will choose Michigan.

In addition to predicting the future, the model provided empirical evidence that BCS schools enjoy a prohibitive advantage over their non-BCS brethren in recruiting top talent. It also disproved several long-held beliefs about recruiting. For example, recruits don't seem to care how many players a school puts in the NFL, they aren't as interested in early playing time as they claim and scholarship reductions actually increase the likelihood that a top recruit will pick a particular school.

So what possessed three highly educated professionals to devote countless hours of their spare time to predicting the whims of 17- and 18-year-old football players? The easy answer is they live in the South. The trio met in the early '90s as doctoral candidates at Florida State. DuMond and Lynch, both Floridians, love the Seminoles. Platania, meanwhile, roots for her home-state West Virginia Mountaineers. And even though they had scattered -- DuMond works for a Tallahassee, Fla., consulting firm, Lynch teaches economics at Mercer in Macon, Ga., and Platania teaches at Elon in North Carolina -- they still yearned in 2004 to understand why certain prospects chose certain schools.

"You read interviews with some of these recruits ... and they say 'I felt more comfortable there' or something really vague like that," DuMond said. "We were just trying to figure out if we could put any science behind that sort of decision."

So DuMond, Lynch and Platania scoured the archived data on for the recruiting classes of 2002-04. During an 18-month span, they devised a set of more than two dozen variables (official visits, distance from the player's home, school recruiting budgets, age of each school's stadium, etc.) and built a model. They devised equations that told them what mattered most when a recruit made his decision. After plenty of heavy thinking and trial and error, they developed a statistical formula they believed would accurately predict a recruit's college choice.

Using a computer program called SAS (Statistical Analysis Software), the trio fed the pertinent data for the 2005 Rivals 100 into the model. They ignored whether players had already committed, instead using each player's final set of schools to see if the model would spit out a correct prediction. The model went 71 for 100, and it correctly guessed the destinations of the six highest-ranked players (Derrick Williams, Jerrell Powe, Eugene Monroe, Fred Rouse, Callahan Bright and Mark Sanchez).

Early in 2006, the trio presented its findings at an economics conference in New York. There, the economists reported that even after controlling for success -- throw out the megapowers -- a recruit is still almost six percent more likely to pick a BCS school over an otherwise equal non-BCS school.

"If you are a BCS school ... you still get a little bit of an additional bump when it comes to trying to land the big players," Lynch said. "It's kind of the rich get richer is what we showed in the development of this model."

But as recruitniks, the trio considered some of the paper's other findings more fascinating. Not surprisingly, they discovered that distance from a player's hometown and whether the recruit made an official visit were the two most important factors, but distance is less of a factor depending on where the recruit lives. With help from geographer buddy Juan del Valle, who used geographic information system (GIS) software to determine straight-line distances between hometowns and campuses, the economists learned that recruits in the South and the Midwest are likely to stay close to home, while recruits in the West and the Northeast seemed more willing to leave the nest.

More surprising were the factors that didn't seem to affect a prospect's decision. "There were a couple of results that had us raising our eyebrows," Lynch said.

DuMond and Lynch each pointed to the roster depth variable. Recruits often mention early playing time as a contributing factor in their decision, but the model found that prospects are one percent more likely to choose a school that a year earlier signed one or more highly touted players at that recruit's position. That could explain how USC keeps signing running backs. Meanwhile, scholarship reductions didn't seem to bother recruits. The economists surmised that the players figured they'd have less competition for exposure and playing time. Also surprising, graduation rate and the number of players sent to the NFL in recent years had no measurable effect on recruit's choices.

So what do recruits want? According to the model, they usually will pick the BCS-conference school nearest their hometown that has the biggest on-campus stadium and won the most games last season. Not the past five seasons, mind you. "You can make the argument that recruits may be a little short-sighted," DuMond said.

DuMond and his partners, however, intend to take the long view. Lynch said there are plenty of college football fans in the economist community, and many have suggested ways to tweak the model to make it more accurate. The trio will keep working, Lynch said, but the complexities of the teenage psyche will keep them from reaching 100 percent accuracy.

"At the end of the day, they're just kids trying to figure out where they want to go play ball," Lynch said. "We'll be able to accurately forecast a bunch of them, but I don't think a day is going to come where we're going to accurately pick every one of these kids."