“What Works” in Education Is Not Merely A Question of Effect Sizes

Here’s a pet peeve: A champion of some particular education intervention will point to some research study showing Intervention X led to positive outcomes for participating students. Ok, great. Assuming it’s a good study, we now have evidence that Intervention X “worked” in a given Situation Y.

That does not mean Intervention X will easily replicate to new Situation Z.

Anytime you hear someone attach the phrase “high-quality” in front of some intervention, they’re talking about this problem.  These caveats pop up frequently in debates over particular reforms:

  • The small school reform effort produced long-term gains for students, but their backers largely abandoned them.
  • School integration efforts produced large gains for black students (with no harm to white students), but formal integration programs were and remain relatively small.
  • “High-performing” charter schools produce large gains for students, but there is wide variation in those results.
  • Teacher evaluation reforms produced gains in some cities, but the effects were smaller or non-existent when similar reforms were spread more broadly.

Some of these reflect implementation challenges. Others are more about politics (which is itself a particular type of implementation challenge). I am by no means the first person to make this point, but we can’t just say something “works” or “doesn’t work” without giving some consideration to where the policy worked, for whom it worked, what outcomes changed, and by how much it changed the status quo.

–Guest post by Chad Aldeman

Absolute Versus Relative School Segregation

“Everyone knows” school segregation is getting worse, right? Well, no. That narrative has been fueled by a partial misread of the data.

It is true that absolute measures of school segregation are getting worse, but that’s mainly due to our diversifying country. In relative terms–how racially isolated are our schools compared to the underlying student population–segregation has not been nearly as dramatic. Here’s Brian Kisida and Olivia Piontek in a new piece over at Education Next:

In contrast, relative measures of segregation take into account the underlying composition of students, making them more comparable across locations and over time. They are also conceptually different in that they measure how evenly a given population of students is distributed across an entire school system. This makes intuitive sense, as segregation implies that some students are segregated from other students—relative to some underlying pool of students a school could enroll.

This issue with measuring segregation is well-known in the academic community, and there is ample scholarly evidence using relative measures of segregation that adjust for the underlying composition of students in school systems. Using these more sophisticated relative measures, such as the dissimilarity index and the variance-ratio index, examinations of trends find that segregation has been flat or modestly decreased over the past 20 years. In summary, massive resegregation is not occurring, and students are roughly just as evenly distributed across school systems as they were 20 years ago.

To be clear, segregation and racial isolation are serious problems in America and far more common than they should be, and efforts to shed a light on these problems are commendable. At every point in time between the Brown decision in 1954 and 2019, millions of American children have been educated in separate and unequal schools. This is an outrage that demands serious attention and action. But as we devise strategies and search for solutions, it is imperative that we are motivated by a complete picture of the problem we are trying to address. Segregation is a serious enough problem that it shouldn’t need to be worsening to be alarming. It’s bad enough as is.

–Guest post by Chad Aldeman