At the risk of sounding glib, I was asking for data and not anecdotes. >The mistake from a scientific point of view is how to define legacy preferences and how

At the risk of sounding glib, I was asking for data and not anecdotes. >The mistake from a scientific point of view is how to define legacy preferences and how to measure such impacts. The paper looked to define legacy preferences using multiple datasets where the school measured the importance of alumni relations in admissions. The datasets had to agree for a legacy admission classification (e.g., both say that alumni relations are "very important" regarding admissions). The measure used in the studies that showed no evidence was the level of alumni donations. It's pretty easily quantifiable. Other studies cited show that when legacy admissions were abolished there was not a statistically significant change in alumni donations.