Education researchers have paid a lot of attention to the sorting of teachers across schools. For example, it is well known that schools serving more low-income students tend to employ teachers who are, on average, less qualified (in terms of experience, degree, certification, etc.; also see here).
Far less well-researched, however, is the issue of sorting within schools – for example, whether teachers with certain characteristics are assigned to classes with different students than their colleagues in the same school. In addition to the obvious fact that which teachers are in front of which students every day is important, this question bears on a few major issues in education policy today. For example, there is evidence that teacher turnover is influenced by the characteristics of the students teachers teach, which means that classroom assignments might either exacerbate or mitigate mobility and attrition. In addition, teacher productivity measures such as value-added may be affected by the sorting of students into classes based on characteristics for which the models do not account, and a better understanding of the teacher/student matching process could help inform this issue.
A recent article, which was published in the journal Sociology of Education, sheds light on these topics with a very interesting look at the distribution of students across teachers’ classrooms in Miami-Dade between 2003-04 and 2010-11. The authors’ primary question is: Are certain characteristics, most notably race/ethnicity, gender, experience, or pre-service qualifications (e.g., SAT scores), associated with assignment to higher or lower-scoring students among teachers in the same school, grade, and year? Although this analysis covers just one district, and focuses on a specific set of student and teacher characteristics, it’s a big step forward.
Needless to say, one should not examine the association between student performance and any of these traits in isolation. For one thing, they are all interrelated– e.g., if male teachers, on average, are more experienced than their female colleagues, differences between male and female teachers in the performance of their assigned students may be partially due to the underlying differences in experience (see this 2006 analysis of within-school sorting in North Carolina). In addition, looking at school- and classroom-level factors, such as student characteristics or accountability pressure, may provide additional insight regarding contextual variation in assignment processes.
Accordingly, the authors of this paper fit multiple models using both administrative and survey data to try and unpack some of these relationships. Theirs is probably the most detailed analysis available on this topic.
In the models using only the administrative dataset (i.e., just race, gender, experience, and degree), they find that less experienced, female and minority teachers are more likely to be assigned to classes with lower-scoring students vis-à-vis more senior, white male colleagues, all else being equal. These results are substantively important and interesting, and have been covered elsewhere (see Stephen Sawchuk’s well-done article). But there’s quite a bit more to these relationships than the “overall” findings.
With regard to the “gender gap” – i.e., the difference in test-based performance between students assigned to male versus female teachers – the estimated discrepancy is really quite small, and basically “disappears” when one accounts for the greater presence of female teachers in special education. In other words, special education students tend to score a bit lower on tests, and female teachers are more likely to specialize in special education; so, when the authors control for this, the small estimated gap is erased (actually, reversed).
Discrepancies between white and minority teachers and between more and less experienced teachers are substantially larger. For instance, on average, black and Hispanic teachers are assigned to classes with more lower-scoring students than are white colleagues teaching in the same school and grade. The magnitude of this difference is educationally meaningful. A large portion of this gap, however, is “explained” by the fact that minority teachers tend to be assigned more poor and minority students, and these students tend to score lower than white students from higher-income families (see this paper for more on this phenomenon). Controlling for classroom composition reduces the estimated gaps to the point where they’re rather modest.
And the differences are further reduced when additional teacher characteristics are added to the model (using the survey dataset), including SAT scores, the selectivity of their undergraduate institutions, and whether teachers ever served in leadership roles. When all these factors are included in the model, assignment differences by teachers’ race/ethnicity are mostly explained away.
Finally, it’s definitely worth noting that these gaps are larger in schools with white (versus minority) principals, and that lower-scoring students are more likely to be assigned to black teachers who work in schools with a larger proportion of white teachers.
The findings for experience require similar disentangling. The relationship between years of service and student performance is positive – i.e., on average, higher-scoring students are assigned to more senior teachers. Some experience-based gap persists in basically every specification the authors report, but it does vary in magnitude.
Teachers with between 2-7 years of experience in the district are assigned students who, on the whole, do not score discernibly higher than those assigned to first-year teachers. By contrast, the scores of students assigned to teachers with between 10-20 years on the job tend to be substantially higher vis-à-vis the students of first-year teachers, even when the relationship is modeled as non-linear, and also when the researchers follow the same teachers over time to see if the associations are “real” and not just a result of differential attrition. This is consistent with the limited prior research on within-school sorting – e.g., this analysis of data from Florida and this study, which uses a nationally-representative dataset.
There are three interesting notes about these experience-based gaps. First, they are not larger in middle and high schools compared with elementary schools, which is counterintuitive insofar as “tracking” is generally seen as more prevalent in the former versus the latter. Second, the relationship between experience and student performance is significantly stronger in schools with more senior teachers, but weaker in schools where larger proportions of students are not proficient (i.e., those that may be experiencing accountability pressure, which may influence classroom assignments). Third, there is some indication in the results that experience in the district is less important than how long teachers have served in their specific school.
So, overall, there does appear to be systematic sorting of teachers within schools by measurable characteristics such as gender, race/ethnicity and experience (as well as, by the way, whether or not teachers have ever served as a professional development instructor and the selectivity of their undergraduate colleges).
This analysis is descriptive, and there are a variety of possible ways to explain or interpret its many findings. On the whole, there appears to be a rather complicated web of factors underlying teacher assignment, including teacher, principal and school characteristics. In other words, teachers and principals (and, in many cases, perhaps parents as well) consider a bunch of different factors when sorting students into classes, and this process also varies by context (e.g., by school).
For instance, as the authors note, the fact that more senior teachers are given classes with higher-scoring students may indicate that advanced students are being assigned to teachers with the stronger content knowledge necessary to teach advanced classes, whereas, in other cases, it may be a result of teachers accruing influence and building relationships as they spend more years in the same school.
In any case, efforts to understand what’s going on with classroom assignments bear directly on issues, such as teacher turnover and the use of value-added estimates in evaluations and other personnel policies. As just one example, the fact that novice teachers tend to be assigned lower-scoring students may exacerbate new teacher attrition and mobility. Analyses such as this one are therefore important, and will hopefully spur further examination, especially given the fact that new evaluations and other policies (e.g., ramping up school accountability systems) may influence the teacher/student matching process going forward.
- Matt Di Carlo