Weekend greetings, CharterFolk.
I got up this morning planning to write on another topic and then saw a flurry of pieces that made me change my plans.
The first was the Washington Post Editorial Board …
… following up on Eva Moskowitz’s op-ed from last week …
… excoriating New York for simultaneously backing away from longstanding measures of academic performance and refusing to release results that would allow the public to understand how schools are actually performing.
At the very same moment that understanding the skill levels of young people independent of AI systems is becoming paramount – so that we can know whether our teenagers are engaging in cognitive offloading – society is choosing to cognitively offload our own responsibility to know.
It’s utter lunacy.
In the weeks ahead I will be writing about some new regressions I have been running with the assistance of Astra.
It’s mind-boggling what Astra can help you do.
Below you’ll see me provide something of a sample, but for now I’ll make this basic point.
In terms of the quality of data allowing for the quality of regressions – and yes, to an important degree, the quality of understanding – we are in a strange place.
For grades 3–8, many states give us reasonably consistent statewide assessments. The regressions aren’t perfect. Far from it. But the R-squareds can be respectable enough that we have a meaningful statistical window into what is happening.
We kind of have an understanding.
Then we get to high school.
Exactly the years that matter so much right now.
We have vast swaths of the country with no meaningful data for students in 9th and 10th grades. Other states report such a morass of gobbledygook that even when a regression comes back with a decent R-squared, you discover that the underlying measures tell you very little about what makes schools work.
Or think about the other end.
By third grade, the variation in skill levels that kids and schools encounter is already enormous.
Where did it come from?
Before kindergarten? Or since the child enrolled?
Which schools significantly changed student trajectories?
It’s not that we have poor R-squareds.
It’s that, in most places, we don’t have comparable statewide academic outcomes from which to calculate R-squareds at all.
No common data.
Utter dark-flying.
Which brings me to the article that finally tipped me over the edge and caused me to write on something different today.
Nikole Hannah-Jones’s article in the New York Times this morning …
… revisits the themes she first laid out in her equally autobiographical 2016 rendering …
… of her effort to find a school for her daughter that would both do right by her daughter and satisfy her desire to participate in something that might make schooling better for all kids.
The new piece contains an excruciating moment.
Her now-16-year-old daughter asks whether her parents knew her school was “bad like this” when they decided to send her there.
That question hangs over the entire essay.
On one level, it is an extraordinarily brave article.
On another, I found it just incredibly sad.
And I think the tension between those two reactions begins with something almost every parent can understand:
Cognitive dissonance.
Parenting may be one of the places where cognitive dissonance is most powerful.
We make choices for our children believing – needing to believe, really – that we are doing right by them.
And then evidence begins to accumulate that those choices may not have served our children well.
To absorb that evidence, you have to hold two painful ideas in your head at the same time:
I am a good parent who tried to do what was best for my child.
And:
One of the most consequential choices I made for my child may not have given her what she needed.
In Hannah-Jones’s case, the tension is made even more resonant because part of her parental motivation was trying to watch out for her own child and all children simultaneously – especially kids who need greater fairness most.
Andy Rotherham has made a related observation for years about parent surveys of schools: admitting that your child attends a suboptimal school raises an uncomfortable question about why, as a good parent, you continue sending your child there.
That doesn’t make parents dishonest.
It makes them human.
Which is why I have enormous empathy for Hannah-Jones.
Her daughter is forcing precisely that confrontation into the open, and Hannah-Jones puts it on the page.
But the article also shows how difficult it can be to follow cognitive dissonance all the way through.
Because amid a very long examination of segregation, attendance boundaries, selective admissions and the ways families with means secure better educational opportunity, one word never appears:
Charter.
How does a source as knowledgeable and well-informed as Hannah-Jones withhold from her rendering the fact that New York City has another large public-school community operating alongside the district – one that does not academically screen students, does not guarantee admission based on residence, and yet generates dramatically stronger results?
Her explanatory framework emphasizes forces that are enormous and real and that infuse our movement with purpose:
Segregation.
Racism.
Attendance boundaries.
Selective admissions.
Generations of unequal opportunity.
But what happens when another set of public schools serves many of the same historically underserved students and produces radically different outcomes?
Before I answer that, let’s look at the data.
The analysis below uses results from the 2024–25 school year and asks a straightforward regression question:
Controlling for observable demographics of the students served, do schools’ actual results exceed or fall short of prediction?
I then place those residuals – actual minus predicted – into performance bins, from greatest underperformance on the left to greatest overperformance on the right.
I limited the analysis to grades 3–8, where New York provides common statewide ELA and math assessments that make genuinely comparable school-level analysis possible.
The statewide models control for race and ethnicity, economic disadvantage, English learner status, students with disabilities, grade and relevant interactions.
The cross-validated R-squareds come in at about .41 for ELA and .40 for math.
Not perfect.
But useful.
And what emerges in New York City is striking.
Here is the combined ELA/math distribution for all students broken into quartiles.
The first thing I’ll point out is that district performance is not an absolute train wreck.
That’s actually important, and something we have to recognize when it occurs.
When I ran similar methodology in Indianapolis, the district distribution was much more heavily concentrated toward the underperforming side.
That isn’t New York City.
NYC district schools overall skew somewhat toward the overperforming side of the statewide distribution.
But now look at the charter distribution.
Approximately 62% of charter tested-student records fall into the statewide top adjusted-performance quartile, compared with about 37% of district records.
Only about 7% of charter records fall into the bottom quartile.
Imagine what the state of learning in New York City would look like if dramatically more students had access to schools producing outcomes like that.
Now consider who the two sectors serve.
Charter schools serve about 19% of all grades 3–8 students in this universe.
But roughly 36% of all Black students.
And economically disadvantaged and Black students both make up larger shares of the charter population than of the district population.
Now look specifically at economically disadvantaged students.
Roughly 61% of economically disadvantaged charter students are associated with school-grade units in the top adjusted-performance quartile, compared with about 33% of economically disadvantaged district students.
And then look at Black students.
Here the divergence becomes enormous.
Approximately 64% of Black charter tested-student records are associated with the statewide top quartile of adjusted performance.
For district schools, the figure is about 28%.
Meanwhile, only about 6% of Black charter records fall in the bottom quartile.
Break the distribution into tenths and the difference becomes even easier to see.
About 43% of Black charter records fall into the statewide top adjusted-performance tenth.
The district figure is about 13%.
And fewer than 2% of Black charter records fall into the bottom tenth.
Or look at it in estimated numbers of students.
The distribution implies roughly 15,000 Black charter students associated with schools in the statewide top adjusted-performance tenth, compared with roughly 8,000 Black district students.
District schools serve substantially more Black students overall.
Yet charter schools account for nearly twice as many Black students associated with schools in the highest adjusted-performance tenth.
And lest anyone worry that these results merely reflect relative overperformance against alarmingly low absolute proficiency, consider the raw results.
In 2024–25, NYC charter students posted raw proficiency rates of roughly 68% in ELA and 69% in math, compared with roughly 56% and 57% in district schools.
Among Black students, the differences were much larger: roughly 68% charter versus 47% district in ELA, and about 69% versus 43% in math.
There is an entire public-school movement in New York City serving very large numbers of Black and economically disadvantaged children, without academically screening them, and producing dramatically different academic results.
And somehow that movement does not merit a single mention in an article centrally concerned with educational opportunity for Black children in New York City – and with the history of a system that has repeatedly sorted access to better opportunities toward those with means and worse opportunities toward those without.
Particularly because the article is so honest about Hannah-Jones’s daughter experiencing school conditions that nobody should accept as inevitable, it is all the more striking that it omits another public-school sector demonstrating that those conditions are not, in fact, inevitable.
There are things schools cannot wish away – segregation, poverty, centuries of racial injustice.
But weak academic programs, institutional disorder and adults failing to perform at reasonable levels of competence are not immutable consequences of an unjust society.
We know that because other schools serving similar children are producing profoundly different outcomes.
Which means parents do not necessarily have to choose between securing a great education for their own children and working toward a system in which every family eventually has access to one.
We can enroll our own kids in excellent schools that do not screen others out while pushing to make that opportunity available to everyone.
And that is where New York City should make all of us profoundly uncomfortable.
Because this city has helped birth one of the most consequential public education reform efforts in the country at scale.
Charter schools serving large numbers of Black and economically disadvantaged children are producing dramatically stronger results.
And here is the incredible irony:
The very schools generating those results are effectively subject to a moratorium on meaningful growth under New York City’s charter cap.
Families want more of them.
And yet our public policy response is to prevent enough of those schools from opening to meet demand.
It’s a collective refusal to confront the R-squareds amassing before us.
Not one article.
An entire societal conversation increasingly structured to avoid confronting things we know to be true.
Across more cities, more states and more analyses, we are accumulating enough evidence to see recurring patterns – including a critical mass of schools serving children who historically have been denied great educational opportunity and producing outcomes far beyond what demographics would predict.
That does not justify some fantasy-laden zeal that we have figured everything out.
It justifies something far more modest – and far more important:
A general sense of confidence that better is possible.
Enough confidence to believe that, if we act with courage, better outcomes for thousands upon thousands of children – and perhaps ultimately millions – wait on the other side.
And when the discomfort created by that realization becomes so great that we push the evidence out of view – when the R-squareds are sitting in front of us and we refuse to grapple with what they reveal – that is when cognitive dissonance becomes cognitive offloading.
Offloading the responsibility to see.
The responsibility to reconsider.
The responsibility to change.
None of this diminishes the courage or heartbreak in Hannah-Jones’s piece.
Her willingness to put such a painful family reckoning before the public should deepen our empathy for how difficult it is for any of us to confront evidence that unsettles beliefs bound up with choices we have made for our own children.
But that is precisely why the lesson matters.
As we grapple with the unknown potential future of cognitive offloading in an AI world we can barely imagine, we should take heed of the cognitive offloading we already know how to avoid.
The kind that happens when evidence is in front of us and we choose not to see it.
That is the cognitive offloading we can stop right now.