When public school students opt out of standardized tests, does it really matter? Sure, the resulting scores will be based on a smaller statistical sample, but does that have any bearing on the real problem of educational inequity?
It's natural for parents to believe they're better-qualified than anyone else to make decisions for their child. That said, it's also true that, every day, people become parents who would be turned down if they tried to adopt a dog at the Humane Society.
One prominent example of parental certainty is the opt-out movement alluded to above, where concerned parents are yanking their kids out of the Common Core-aligned Smarter Balanced Assessment test in growing numbers.
Putting aside the ample political controversy over this practice (see "Cheating on Tests," WW, Oct. 21, 2015, if you're into that sort of thing), what about the science? Does opting out change aggregate test results? A recent data simulation from the Brookings Institution suggests the numbers can take a surprising amount of abuse.
As Carlos implies, if kids opt out at random—with low scorers just as likely to bag the test as high scorers—the data get less precise, but they don't acquire an erroneous slant.
However, that's not what's happening. Like soy curls and Infowars.com, opting out appears to be more popular among wealthier families, whose kids tend to score higher than average. Even with that skewed sample, though, nearly 20 percent of a given cohort need to opt out before the data really starts to collapse.
Unfortunately for fans of testing, 13 of 96 Portland Public Schools are already at or beyond that point, and opt-out sentiment is not exactly fizzling.
All of the above also ignores the fact that federal regulations mandate a 94.5 percent participation rate in the test. PPS is currently about 87 percent overall, so the U.S. Department of Education could theoretically yank our Title I money any time. (Humane Society funding, thankfully, would remain unaffected.)
Questions? Send them to firstname.lastname@example.org