This principle helps explain why people behave differently from what pure logic would predict.
More quotesRanking measure: acceptance rate.
Response: recruit more applications, including from students unlikely to be admitted.
Measured result: acceptance rate falls. Selectivity appears to rise.
Applicant result: more people spend time, fees and hope on an application with little chance.
Ranking measure: proportion of small classes, with the threshold at twenty students.
Response: cap classes at eighteen or nineteen.
Measured result: more small classes.
Teaching result: a room of nineteen qualifies; a room of twenty misses the cutoff.
Ranking measure: average SAT score.
Response: exclude some lower-scoring groups from the requirement, admit others in periods outside the count, pay admitted students to retake the test.
Measured result: average rises.
Corrupt people are unnecessary. The target needs ordinary people who understand what receives reward. Each local action can be defended. Together they detach the score from the quality it was meant to summarise.
A wider bundle of measures creates more surfaces for gaming. A single elegant number creates one obvious surface. Neither arrangement abolishes incentives.
The next ranking release shows improvement.
Applications from unlikely candidates have also risen. Their outcomes occupy no column in the published score.
The next ranking release is scheduled for September. Data collection closes on 31 July. Publication follows eight weeks later.