Individual Value Added (IVA) is the go-to method for schools to determine your statistical worth as a teacher. But is IVA reliable enough to give a precise score on how effective of a teacher you are?
In reference to Individual Value Added (IVA), DCPS states teachers are not allowed to calculate their own IVA score because “calculating a likely score involves a complex statistical process known as regression analysis”. Any teacher with a quantitative background would know that regression analysis is statistics 101, not a formula beyond the comprehension of mere mortals. I calculated individual value added scores for my take-home final in my economics of education class in college, and it was by far the easiest question on the final. If they released the dataset, it would not be hard to calculate IVA and confirm their teacher ratings. However, DCPS knows that if any third party did IVA calculations, the ugly truth would come out: statistical noise is too large for the stakes of IVA to be as high as they are. IVA is quite frankly not precise enough for peoples’ livelihoods to hinge on.
Statistical noise is made up of the unknown factors still influencing our data that our model cannot possibly capture. With even the best social science models, the “r squared’, or the amount of variation in the data that can be explained by our model (in this case, IVA), rarely exceeds 50%. When it does, social scientists begin doing cartwheels. This means that the margin of error on the true “value added” score is most likely very high, especially for teachers with small student sample sizes. Elementary school teachers should be even more worried than secondary ones, because their sample size cannot possibly be big enough to be precise.
Stephen Raudenbush, one of the world’s leading experts in how to statistically evaluate the effectiveness of schooling, best put it in this article that IVA is far too statistically noisy to truly figure out how is the “best” teacher. With 95% confidence, we see that the majority of teachers are statistically indistinguishable from one another. The outliers, Teacher B and Teacher A, we can at least definitively say did or did not add some value to their student’s learning. But we cannot really precisely say how exceptionally good or exceptionally bad they truly are.
You can say with confidence that Teacher A did not teach anything to their students and should be fired. And you can be certain that Teacher B did teach something to their students, even if you’re not certain how much. For 95% of teachers, however, you can’t say much of anything at all. The fact that this score accounts for 35% of teacher evaluations, and determines if teachers get upwards of a $25,000 bonus, is absurd. It causes large amounts of stress in the teachers and is not precise enough to make a definitive decision on exactly how good or exactly how bad a teacher really is. Yet, DCPS hides behind the phrase “complex statistical formula” and pretends that we can.