Some semi-random thoughts on randomization

Paufler and Amrein-Beardsley’s 2013 article “The random assignment of students into elementary classrooms: Implications for value-added analyses and interpretations” analyzed the impact of the randomization and non-randomization of assignments of students in Arizona classrooms. This response will attempt to highlight and describe insight found in the article.

First off, I’d like to start with a thought on randomization of students in classrooms. This premise of randomized classroom groupings begins with the conscious pairing of students by their ages. In my mind, a truly randomized classroom in, for instance, a K-8 school would feature six-year olds and their teenage counterparts in the same classroom. So, the students studied for this would be randomly grouped once they are divided in grade levels based on their age. Paufler and Amrein-Beardsley (2013) wrote this regarding randomization, “The purpose of random assignment is to make the probability of the occurrence of any observable differences among treatment groups (e.g., treatment or no treatment) equal at the outset of any experiment or study” (p. 3). This type of equality is important when the design is to study the impact of randomization. In terms of probability, all classrooms would have equal chances of having a wide-variety of students within them, running the gamut from students with disabilities to gifted learners. Studied over time in aggregate, randomization would precipitate and further comparable numbers of all types of students. Paufler and Amrein-Beardsley (2013) addressed issues regarding thoughts of randomization in classroom prior to and post standardized tests, “Whether students have been randomly assigned to schools and classrooms has not mattered much in the past because until recently, teachers were not typically held accountable for the test scores their students attained (i.e., once per year on traditional ‘‘snapshot’’ standardized tests)” (p. 8). Now that we are in the era of standardized tests, it matters the types of students that teachers receive; it matters as well the means that teachers receive their students. It’s one thing if each teacher receives an equal number of special education students in his or her classroom – this may just be a product of the population. It’s a completely different issue if one teacher receives ALL the special education students in his or her classroom because he or she is “good” with them and then is assessed with the same metrics of standardized tests that his or her colleagues are per grade level. I can see these non-randomized groupings happening all the time pre standardized tests. Something like: Shelly takes all the autistic students because she’s so patient with them, or Sandra takes all the gifted students because she knows how to challenge them. I can foresee scenarios where hypothetical teachers like Shelly and Sandra enjoy being these people on campus but not when they now are assessed as teachers for something like a pay bonus against their colleagues who do not have these special needs students in their classes. With a raise on the line, Shelly and Sandra may no longer request students they enjoy teaching. They may mandate a randomized classroom selection model.

Analyzing scenarios such as the aforementioned through the lens of impact, one can see how standardized tests greatly affect a teacher’s view of his or her class roster and how it is selected. Paufler and Amrein-Beardsley (2013) went on to discuss student growth scores and their linkage to teachers, “While it is certainly reasonable to ask to what extent bias occurs given varying student placement practices, in the state of Arizona, students’ growth scores are still not linked to teachers’ records to permit teacher-level value-added analyses for such a purpose” (p. 12). It seems as if the threat (and threat is too strong of a word) of test scores continually on the horizon might make it impossible to focus upon or assess a purely randomized classroom structure. The stakes might just be too high for principals not to front load their perceived stronger teachers with students deemed troublesome. All of this having been said, the attempt is both noble and interesting. Whether or not classrooms are truly randomized is outside the scope of the study. The initial premise was to study randomized classrooms against non-randomized classrooms. Non-randomization has been going on for years. Teachers, administrators, and parents have “cherry-picked” classrooms based upon pre-conceived notions of students and teachers. This is less and less equitable as standardized tests become more prevalent.

Paufler, N.A. & Amrein-Beardsley, A. (2013). The random assignment of students into elementary classrooms: Implications for value-added analyses and interpretations. American Education Research Journal, 51(2), 328-362.