KDnuggets : News : 2003 : n18 : item16 < PREVIOUS | NEXT >


Data mining by economist revealed cheating in Chicago schools


Steven D. Levitt, 36, a professor of economics at the University of Chicago, achieved tenure after only two years.

Called the "most brilliant young economist in America" in a recent New York Times Magazine article, Levitt is known for his flair for finding creative answers to questions that wander far beyond the boundaries of traditional economic theory.

Levitt set off a controversy by concluding that the easy availability of abortion in the 1970s helped cause a sharp drop in crime two decades later.

Guided by a suspicious mind, as well as an instinct for deciphering the hidden ways of the world, Levitt chips away at mountains of data to unlock the secrets behind systems of cheating -- from horse racing to real estate sales.

One of his most celebrated papers looked at the distribution of wins and losses among Japanese sumo wrestlers going back 20 years. The odd patterns of victories that Levitt discovered helped him to pinpoint which wrestlers probably won matches by bribing their opponents.

The statistical methods Levitt developed in his study of sumo wrestlers helped him devise the algorithms that enabled him and Brian A. Jacob, an economist who teaches at Harvard University's Kennedy School of Government, detect cheating among teachers in the Chicago public school system on the Iowa Tests of Basic Skills.

Their system for spotting cheaters eventually led the Chicago school system to fire four teachers and a teacher's aide and to issue letters of reprimand to two principals.

The teachers were accused of misconduct that included erasing incorrect answers, pointing to correct answers during the test and filling in answers to questions that students had left blank.

Although the Levitt/Jacob system for spotting cheaters provided a powerful detection tool, it could not by itself prove that a teacher had actually cheated.

To reach that point, Chicago school officials followed an elaborate path of verification.

"We basically use their data to narrow the focus," said Dan Bugler, the Chciago schools' chief officer for research evaluation and accountability.

Students in classrooms suspected of cheating retake the Iowa tests. Their test results then are compared with the test results achieved by control groups of students in classrooms not suspected of cheating, Bugler said

If results still look odd, the investigators will scrutinize the original tests for unusual patterns of erasures or answers, Bugler said.

"They might see a classroom where 20 kids have identical answers on the last 10 questions of an exam, for example," Bugler said. "And at that point we turn it over to the legal department and they interview kids and interview teachers."

KDnuggets : News : 2003 : n18 : item16 < PREVIOUS | NEXT >

Copyright © 2003 KDnuggets.   Subscribe to KDnuggets News!