Scoring Outliers' Effect on Race to Top
Louisiana officials are complaining that one Race to the Top judge sank their entire application.
Well, you can judge for yourself, but our own quick analysis shows Louisiana was affected more than any other finalist state by scoring outliers.
With the help of Stephen Sawchuk of Teacher Beat fame, we cobbled together our own scoring spreadsheet, throwing out the lowest and the highest score for each state, and averaging the remaining three scores. This means that a really hard grader, or a really easy grader, can't unduly influence the scores. We'll call this the McNeil-Sawchuk scoring system, inspired by the various scoring system changes in international figure skating.
The new scoring system would not have changed the outcomes for Delaware and Tennessee, which would have still been Nos. 1 and 2, respectively.
However, this would have vaulted Louisiana by 10 points to sixth from 11th place (and higher had it not lost those easy 15 STEM points), and Georgia would have fallen by 6.6 points from third place to seventh.
It turns out, Georgia and Louisiana were most affected by outliers. Georgia benefited from an unusually high score (jump to Page 29 of the state's comments to read why that judge thought so highly of Georgia). And Louisiana suffered because of a low one (jump to Page 31 to read that judge's comments).
The McNeil-Sawchuk scoring system has its many flaws, especially since there might be very good reasons for such big swings in judges' scores. But it's important to acknowledge the power of a single peer reviewer whose unusually high or low score could have a big impact on states' rankings.