Daniel Koretz of Harvard's Graduate School of Education took the lecture notes from his course, "Methods of Educational Measurement," and turned it into a book: Measuring Up: What Educational Testing Really Tells Us. It's readable, filled with funny anecdotes, and contains absolutely nothing that will be new to regular GNXP readers.

But because Koretz takes the math and most of the controversy out of the debate over standardized tests, he has time to actually drill home a couple of important points repeatedly: Modern standardized tests have little bias, are pretty reliable, and while they don't tell you everything about a person or a school or a city, they are good for making rough predictions.

Hence, the title of this blog post: Feel free to recommend Measuring Up as a "baby steps" book for your favorite sociologist or folk guitarist.

Koretz waves his political correctness card early on, letting us know that "IQ [is] just one type of score on one type of standardized test..." and he lets us know about the "pernicious and unfounded view that differences in test scores between racial and ethnic groups are biologically determined." But you already knew he was going to say that, right? And in an unintended parody of blank-slatism, he has a chapter entitled "What influences test scores" that never once mentions genetic factors, even to dismiss them.

Koretz does a great job dodging such troubling questions while focusing on what he really wants to talk about, with solid, candid chapters entitled "Validity," "Inflated Test Scores," "Error and Reliability," chapters that actually do a good job of conveying big ideas about non-experimental social science in jargon-free prose. Kudos to him for doing so.

Treat it as a book on the narrow field of psychometrics and its link to policy, not as a book on the broader field of standardized tests per se and its link to policy: You'll spend a lot less time grinding your teeth.

Labels: Psychology