Outliers matter

In response to The Broad Center’s recent “Hire Expectations” report on superintendent tenure, we received some questions about why we focused our findings on means instead of medians. Now we know what you’re thinking. “Medians? Means? I don’t go on the internet to re-live painful memories of stats class. I go for the celebrity gossip and kitten gifs TBC blog!” But please keep reading. This is a great question to explore.

We do have to go back to your intro to statistics class but just for a quick second to set the stage: Means are what folks traditionally refer to as “average,” adding up all the observed values and dividing by the number of observations, whereas medians denote the exact middle value of the distribution of observed values. In other words, 50 percent of values are higher than the median and 50 percent are below it.

Why does this matter? Means are more heavily influenced by outliers and medians are not. So it makes sense why medians can sometimes be more informative and often a better measure of central tendency when data are skewed — especially if the hypothesis is that outliers are true anomalies.

In Hire Expectations, we report that the average (or mean) ongoing superintendency is close to four years (3.76 to be exact). The median is three years. And we report that the average (or mean) completed tenure of superintendents in the largest 100 districts is six years (6.16 to be exact). The median is five years. So why did we focus on means and not medians?

The main reason we made that choice is that our report was published in response to a commonly mis-cited metric about means.

That said, the five-year median is still substantially longer than the three-year figure that is often mis-cited as completed tenure. And among those completed tenures, almost half (44 percent) are longer than the mean of six years that we reported. So the effective shift of using the mean as a measure of center versus the median is not huge.

Additionally, in this work, the outliers matter. Really, the whole distribution matters: means, medians, modes, outliers, spread, skew — all of it! But we can especially learn from the outliers on both ends of the spectrum.

For example, if we just examine medians, it doesn’t matter if the longest superintendent tenure is 25 years or 7 years or if the shortest is 0.2 or 2 years. Either way, the medians stay the same. But it DOES matter that leaders in this space have sustained their large district roles for 10, 15, even 25 years, and it DOES matter that we have leaders in this space who left their roles shortly after they started. We can learn a lot from these leaders — best practices, what to avoid and how to develop and grow.

Leaders who have driven their students and districts to sustained growth, excellence and equitable outcomes have a lot to share. They are #proofpointsofwhatspossible.

At Boston Public Schools, Thomas Payzant served as superintendent for 11 years before retiring as the longest-serving superintendent the school district had seen in more than 30 years. During his tenure, student achievement increased across the board. Math and reading results rose for high school students in all ethnic groups. In fact, the percentage of students passing the state English assessment increased by nearly 80 percent, and the percentage of 10th graders passing the rigorous state math assessment tripled. From 2003 to 2005, fourth and eighth graders in Boston led the nation by making the largest improvement in math on the National Assessment of Educational Progress. Boston consistently outperformed other Massachusetts districts with similar populations, specifically low-income and Black students.

Under Tom’s leadership, Boston Public Schools won the 2006 Broad Prize for Urban Education. It is because of this track record for achieving academic growth in places as different as Boston, Oklahoma City and Eugene, Oregon, that Tom served as a superintendent-in-residence for Broad Academy fellows, mentoring and coaching future superintendents and facilitating Academy sessions for several years.

For the past 16 years, Chris Steinhauser has been superintendent for Long Beach Unified School District, serving nearly 74,000 students. During this time, Chris has led improvements in Advanced Placement enrollment, graduation rates and college preparation — with more students of color taking and passing AP tests, more students graduating as graduation rates increased from 75.9 percent in 2010 to 84.2 percent in 2017 and more graduates meeting course requirements to enroll in the California State University and University of California systems. And since 2009, Long Beach has had a better average daily attendance rate for students than all of California.

This academic growth earned Long Beach Unified recognition as a finalist for the national Broad Prize for Urban Education several times, even after the district won the award early in his tenure in 2003. When he speaks to Broad Residents each year, his drive to continuously improve student outcomes motivates our Broad Residents in their efforts to improve public school systems across the country.

While neither of these leaders are alumni of our programs, they are exceptional superintendents from whom we learn.

It is not lost on us that both of these leaders are White men, which underscores the importance of our findings and hopefully spurs conversations about representation as well as what it takes to expand the pool of talented leaders who are hired for superintendent jobs and create the conditions for success for women and superintendents of color. But in order to do that, we need to start by grounding our collective expectations in the data and paying attention to outliers.

Recent Content

Alumni Guest Series

Alumni Spotlight: Wendy Robinson

Wendy Robinson June 8, 2020