Assistant Professor Iain Pardoe didn't stroll down the red carpet on Oscar night, but he's become a star in the Department of Decision Sciences at the University of Oregon's Lundquist College of Business. Pardoe has perfected a statistical model that uncannily predicts the Oscar winners in the top four categories. This year, his model successfully calculated three of the four winners for best actor, best actress, best director, and best picture.

Pardoe developed the model based on a "discrete choice" economic theory. The model takes into account such criteria as other honors received (for example, The Golden Globes and Screen Actors Guild awards); the number of nominations a film receives; and individuals' past nominations (statistically, it helps if a star's been nominated before, but not if he or she has won). Businesses typically use discrete-choice models to analyze what drives consumers to select one brand's product over another's.

"Casting an Oscar vote has some predictable behavioral characteristics similar to a consumer picking a car. Certain factors matter more than others, and we have to figure out which ones," explained Pardoe.

For this year's Academy Awards, Pardoe's model gave Director Ang Lee a 90.5 percent chance of winning best director; Philip Seymour Hoffman a 94.8 percent shot at best actor; and Reese Witherspoon a 75.9 percent likelihood of garnering best actress. All three won their respective categories.

Pardoe's model also gave Brokeback Mountain a 91.1 per cent probability of winning Best Picture. Although-to almost everyone's surprise-Crash ultimately took home the top honor, such unexpected moments are part of Oscar history, and Pardoe notes that these surprises can provide insights to help refine his model. Nonetheless, his model has been remarkably accurate over time. In fact, Pardoe has run the model against previous Academy Awards, successfully predicting 186 of the past 268 winners since 1938. That's a success rate of 69 percent, and it climbs to 81 percent when only considering awards since 1975.

Pardoe, who teaches statistics at the college, said he originally developed the Oscar model in part as a tool to introduce students to marketing research techniques. "It's a nice teaching example," said Pardoe. "Everyone has an opinion about the Oscars. Students can get excited about the power of statistical modeling when they see the model and its predictions in the classroom."

Pardoe additionally employs similar statistical techniques in his research. Currently, he is collaborating on a study that relies on discrete-choice modeling to investigate how consumer food product choices are influenced by "organic" and "sustainable agriculture" labels.