Caroline Hoxby and Sarah Turner, who teaches economics at the University of Virginia, mailed packages to 40,000 students whose SAT scores and high school grades suggested they were smart but lived in poor neighborhoods, based on census data. They provided individually tailored lists of institutions whose track records indicated they’d be good matches for given students, and information about the college application process, including admissions deadlines and the actual net tuition charged. Once the data were collected, the cost of the experiment came to $6 per student. And the return on ...