The lawyer told the judge that VDOE wanted a try-again on its loss to Brian Davison last year where the judge told VDOE to disclose the SGP data by teacher. VDOE also wanted to bring along the Loudoun School Board, the Virginia School Boards Ass’n, the VEA, and the Virginia Superintendents’ Ass’n. to whine about how terrible it would be to publicly identify the good and bad teachers in Virginia’s public schools.
It looks like the judge divided the baby: He told the various Associations that they were not the affected teachers and they lacked “standing” to intervene in the suit. He allowed the Loudoun School Board to join VDOE in trying to get him to change his mind about disclosing teachers’ identities and, indeed, about releasing the SGP records at all.
It will be up to Judge Hughes to decide the legal questions here. In contrast, it is clear that VDOE is on the wrong side of the policy issue: They are using taxpayer money to resist disclosure to the taxpayers of data those taxpayers paid for. Those data can tell the public which teachers, schools, and school divisions — all paid for by those taxpayers — are doing a good or poor job of public education.
We have a preliminary data release (actually the last of three) that contains anonymous teacher identifiers and that demonstrates the importance of these data.
As a reminder, the Student Growth Percentile measures how much a student has learned in a particular class in comparison to other, similarly situated students. Importantly, the SGP, in contrast to the SOL, is generally unaffected by the wealth or poverty of the student’s family. VDOE has been collecting these data, under a federal mandate, since 2011.
Let’s look at the 2014 Richmond fifth-grade reading data from VDOE’s latest (“16790”) release. For a start, here is the distribution of student score averages by teacher across all the teachers of that subject in all of Richmond’s elementary schools. The SGP percentiles are on the abscissa, the count of teachers with that average percentile is on the ordinate.
Here we see a close-to normal distribution with a mean of 48 and a standard deviation of 13. For comparison, the statewide distribution for this subject (also by teacher average) also averages 48, and with a standard deviation of 11.
For the teacher at “85” on that graph (ID # 66858, average reading SGP of 84.52), the database reports 23 scores. Of those 23 students, only one scored below the state average SGP.
At the other end of the graph, teacher # 66294 averaged only 16.6 but with only ten scores. Let’s look at the next teacher up, # 68809, with 22 scores averaging 22.8:
Three kids in that class scored above the state average. Two scored in the minimum percentile and four more were in the third percentile.
More specifically, the 95% confidence interval of this teacher’s 23 average is 11. In terms of student progress, we can be confident that this teacher is in the bottom third, and probably the bottom quarter, statewide. Clearly it’s time for some retraining and, if that doesn’t take, a replacement teacher.
Your fifth grader in Richmond might be stuck with this teacher. But VDOE doesn’t want you to know how bad this teacher is.
Please recall that the SGP measures improvement in comparison to similarly situated students. A student with high achievement (high SOL score) last year who improves as much as the others with similar achievement, only makes the 50th percentile. A student with low achievement last year and who improves as much this year as the other low achievers also makes the 50th percentile. In short, the SGP does not penalize a teacher for having a bunch of low-performing students; it rewards or penalizes a teacher based on how much improvement that teacher achieves compared with similar students statewide.
Which teacher do you think is doing a better job? Which class would you want your kid to be in?
Why do you suppose VDOE doesn’t want you to know how well (or how badly) your kid’s teacher is doing?
Are you beginning to understand why I refer to VDOE as the State Department of Data Suppression?