Predictive Scores and Ultimate Test Passage

The numeric values selected for this Demonstration are drawn from the relationship between LSAT scores in the United States and scores on the bar examination generally required to become licensed as an attorney in the United States. In recent years, the American Bar Association, whose accreditation of a law school is generally required before that school's graduates may become attorneys, has required as part of "Interpretation 301-6" that pass rates on the bar examination achieve a certain threshold. Various law schools have complained that this standard unduly burdens them because of the lower predictive scores of the students who attend them. Yet, to know whether such a school is in fact doing a good job educating its students to pass the bar exam, one should know the shape of the relationship between predictive scores and bar passage rates. This Demonstration suggests that, as a matter of theory, the shape should be somewhat sigmoidal.

Snapshot 1: lowering the score required to pass the ultimate examination effectively moves the curve leftward

Snapshot 2: increasing the range of predictive scores of admitted students "flattens" out the sigmoidal shape into a more linear one

Snapshot 3: increasing the standard error of prediction (i.e., weakening the correlation between predictive scores and ultimate examination scores) also flattens out the sigmoidal shape into a more linear one

Snapshot 4: shows the relationship when no school admits students whose predictors are so low that fewer than half will pass an ultimate examination