The School Was Rated

Two houses sit on the same street in suburban Boston. Same square footage. Same number of bedrooms. Same year built. Same school district. Same property tax rate.

One sold for more.

The difference between them is invisible from the sidewalk. A school attendance boundary runs between the two properties. The house on one side feeds into one elementary school. The house on the other feeds into another. An economist at Harvard studied exactly this in the mid-1990s, comparing houses on opposite sides of attendance boundaries. She controlled for everything. Same neighborhood. Same tax rate. Same district funding. The only variable was which school the address pointed to. Parents paid 2.5 percent more for a five percent increase in test scores.¹

In a $500,000 market, that is $12,500 for the same house on the other side of the street. In a $2 million market, $50,000. In the major coastal metros where families stretch hardest for schools, the gap is larger still. A Brookings Institution study of the 100 largest metro areas found that homes near high-scoring schools sell for $205,000 more on average than homes near low-scoring schools. Not different homes. Comparable homes. Different school assignments. In San Jose, the premium approaches $500,000. In Los Angeles, $300,000.²

The family checked the number. They compared districts. They stretched the budget to be on the right side of the line. They believe they purchased a better education for their child.

They purchased a number.

The number

At the school level, the relationship between a school's test scores and the income of the families it serves is one of the strongest correlations in education research. In Ohio, the correlation between school poverty rates and school achievement scores is negative 0.79. Nationally, when Stanford researchers analyzed 215 million state test scores representing over 40 million students, socioeconomic factors explained roughly three-quarters of the geographic variation in achievement gaps. The individual correlation between family income and a single student's score is modest. Aggregate to the school level, where ratings operate, and income dominates.³

The College Board publishes SAT scores broken down by family income. The data form a staircase so regular it needs no interpretation.

Four hundred points separate the bottom and top of the income distribution. Every bracket outscores the bracket below it. Every section. Every year. No exceptions. The staircase is not an artifact of a single year's data. It appears in every administration the College Board has published.

Children of the wealthiest one percent are thirteen times more likely to score above 1300 on the SAT than children from low-income families. Roughly 80 percent of high-income children take the test. Seventeen percent score 1300 or above. Among children from the bottom quintile, 25 percent take the test. Two and a half percent clear 1300.⁵

The income achievement gap is now nearly twice as large as the Black-white achievement gap. Fifty years ago, the relationship was reversed. The income gap became dominant not because racial disparities shrank, but because income disparities in educational outcomes grew faster. Between 1974 and 2001, the gap between children at the 90th and 10th percentiles of family income widened by 30 to 40 percent.⁶

The gap exists before the first day of school. Income predicts test scores at kindergarten entry. The gap is already large when children walk through the door. It does not grow significantly through years of schooling. The school did not produce the gap. The gap arrived with the students. The school inherited it. The rating measured it. The rating called it quality.⁶

This is what the number on the screen is primarily measuring. Not how well the teachers teach. Not how much the students learn. How much money the families who live within the school's catchment earn. The number on the screen that convinced the family to pay the premium is, primarily, a measurement of their future neighbors' income.

The loop

The number does not passively observe this. The number produces the sorting that produces the scores that confirm the number.

A school scores high. The rating is published. Families with resources see the rating and move to the district. Property values rise. In the United States, approximately 36 percent of all public school funding comes from local property taxes. Higher property values produce more tax revenue. More tax revenue funds the school. Affluent students arrive carrying more resources from home: tutoring, books, stability, parental education, the quiet advantages that compound before a teacher says a word. Test scores stay high or rise. The rating is confirmed. More families see the rating and move in. The loop tightens.⁷

The school did not become good and then get rated. The rating attracted the conditions that produce the scores that confirm the rating.

White school districts receive $23 billion more in total funding than nonwhite school districts. That is $2,226 more per student, driven primarily by the property tax base underneath them.⁸

A study in Florida tested whether the label itself moves prices, independent of the data behind it. Schools that received an A rather than a B on the state report card saw property values rise even when the underlying test scores at the boundary between the two grades were nearly identical. The letter moved prices. The arbitrary cutoff between one grade and the next produced a measurable premium. The label was doing the work, not the learning behind it.⁹

The inverse is the same loop running in reverse. A school loses families. Rating drops. Property values fall. Tax revenue declines. Funding decreases. Remaining students carry fewer resources from home. Scores drop further. Rating drops further. The families who can afford to leave do. The families who cannot remain in a school the number says is getting worse, in a building the number is making worse. Houses redistricted from higher-scoring to lower-scoring schools lose three to five percent of their value. The label moved. The bricks did not.⁹

GreatSchools.org receives more than 40 million visits a year. Zillow displays the rating on every listing page, reaching over 200 million unique monthly visitors. The school rating is the most widely consulted number in American real estate after the price itself.¹⁰

The rating does not observe quality. The rating produces sorting.¹¹

Three countries

If this only happened in the United States, it would be a policy critique. It happens everywhere.

Outstanding

England's Office for Standards in Education, Ofsted, rates schools on a four-tier scale: Outstanding, Good, Requires Improvement, Inadequate. In 2021, a study in the Journal of Child Psychology and Psychiatry tested what the ratings actually predict. After controlling for prior school performance at age 11 and family socioeconomic status, Ofsted ratings explained less than one percent of the variance in student achievement at age 16. Less than one percent. The average correlation between ratings and student wellbeing was 0.03.¹²

Schools in poor areas are five times more likely to be rated Inadequate than those in affluent areas. Secondary schools with fewer than five percent of pupils eligible for free school meals are three times more likely to be rated Outstanding than schools where a quarter or more qualify.¹³

British parents pay up to eight percent more for a home in the catchment of an Outstanding primary school.¹⁴

British schools are funded by the national government through a needs-based formula called the Dedicated Schools Grant. Local council tax funds waste collection and road maintenance. It does not fund school budgets. The funding loop that exists in the United States, where property taxes flow into school budgets, does not exist in England.¹⁵

British parents are paying for the signal, not better-funded schools. The rating alone, stripped of the funding mechanism, still sorts families by income. The signal is sufficient.

In January 2023, Ruth Perry, headteacher of Caversham Primary School in Reading, took her own life while awaiting publication of an Ofsted report that would downgrade her school from Outstanding to Inadequate. She had led the school for years. The coroner attributed her death to the inspection. In September 2024, the government abolished single-word Ofsted ratings. The label that explained less than one percent of what it claimed to measure had cost a teacher her life before the policy changed.¹⁶

Five point three million

In Beijing's Xicheng district, a room of 11.4 square meters sold in 2016 for 5.3 million yuan. Roughly $810,000. It sat in the catchment of Beijing No. 2 Experimental Primary School. In November 2019, a cubicle of 5.6 square meters, covered floor to ceiling in bathroom tiles, standing room only, sold at auction for 1.28 million yuan. The buyer was not purchasing a home. The buyer was purchasing a household registration that entitled a child to attend the local school.¹⁷

The Chinese call it xuequfang. School district housing. The mechanism is the same one operating in suburban Boston, compressed into a space too small to sleep in. The family is not purchasing square footage. The family is purchasing a household registration number that maps to a school assignment. The space is incidental. The address is the product.

The address determines the primary school. The primary school determines the middle school placement. The placement determines the high school. The high school determines the gaokao score. The gaokao is taken by roughly 10 million students annually, competing for approximately 6.5 million university seats. Fewer than one million of those seats are at top-tier research universities. One address at age six, cascading through a lifetime of sorting. The cascade is what makes the 5.6 square meter cubicle rational. The family is not paying for space. The family is buying a position in a sequence that has already been decided by the time the child turns seven.¹⁸

In July 2021, the Chinese government launched the Double Reduction policy. New for-profit tutoring licenses were banned. Existing institutions were barred from teaching core curriculum. Teachers were rotated across districts starting in 2022, an attempt to equalize the human capital inside the buildings. The intent was explicit: break the sorting mechanism. Prices in elite school districts dipped by two to three percent. In Hangzhou, school district adjustment policies pushed prices in elite districts up more than six percent. Block one channel, the pressure finds another.¹⁹

The government tried to remove the sorting mechanism. The sorting survived.

Twenty billion

Three districts in Seoul, Gangnam-gu, Seocho-gu, and Songpa-gu, hold 43 percent of the city's total apartment value. Property prices in Gangnam run two to three times the citywide average per square meter.²⁰

In 2024, South Korean families spent 29.2 trillion won on private tutoring. Roughly $20 billion. The highest figure ever recorded. Eighty percent of students participate. Daechi-dong in Gangnam, the densest concentration of private academies in the country, has over 900 in a single neighborhood. Families spend more on tutoring each month than on food and housing combined.²¹

PSY's "Gangnam Style" was a satire of wealth signaling in this district. It went viral globally as a dance song. The world danced to a joke about the mechanism.²²

Three countries, three funding models, three political structures, three entirely different educational traditions. The parent in Texas checking GreatSchools, the parent in London checking Ofsted, and the parent in Beijing buying a cubicle are performing the same act. Each one is paying for a number that measures who lives nearby while believing they are paying for education. The mechanism is not a function of property tax policy or school funding structure. The British system proved that by eliminating the funding loop and watching the premium persist. The mechanism is human. Give a parent a number that ranks schools, and the parent will sort. The sorting will produce the scores. The scores will confirm the number.

The teacher

Four times

School A. Students arrive reading at a third-grade level. They leave reading at a fifth-grade level. Two years of growth. The school's rating: low. Students are still below grade level at departure.

School B. Students arrive reading at a seventh-grade level. They leave at seven and a half. Half a year of growth. The school's rating: high. Students are at grade level.

Which school is teaching better?

School A. By a factor of four. The teacher in School A took a child from a third-grade level to a fifth-grade level in a single year. The teacher in School B maintained a child who was already ahead. The rating rewards School B. The rating penalizes School A. The parent who checked the rating chose School B. The teacher nobody celebrates is in the building nobody chose, producing outcomes nobody measures.

In Denver, Knapp Elementary serves a predominantly low-income Hispanic student body. The school scored 9 out of 10 on student growth, measuring how much children actually learn in a year. The same school received 4 out of 10 overall, dragged down by its proficiency score, which measured where students started rather than what the school added. Among 88 Denver schools serving similar populations, one scored above average on the overall rating. Twenty-five scored above average on growth alone.²³

An MIT economist who later won the Nobel Prize studied the gap between the two kinds of measurement. He found that proficiency-based ratings are "particularly off base" and "strongly skewed in ways that hurt schools with more students of color." Growth scores are "much less biased."²⁴

The rating systems that families consult, the numbers that Zillow displays, the scores that drive property premiums, are overwhelmingly proficiency-based. They measure where students are. Not what the school added.

A team of economists at Harvard tracked 2.5 million children from elementary school through early adulthood, linking school records to tax filings. Their finding: replacing a teacher in the bottom five percent with an average teacher would increase the lifetime earnings of a single classroom by approximately $250,000. Not over a career. From a single year of instruction. The same teacher increases college attendance, reduces teenage pregnancy, and raises retirement savings decades later. If earnings are not discounted, the cumulative gains per classroom exceed $1.4 million.²⁵

That teacher is statistically more likely to be in the school with the low number. The building the rating told you to avoid. The teacher producing the most economic value in the education system is in the school with the highest growth and the lowest rating. She is invisible because the number that sorts families was never designed to find her. The number finds income. It always has.²⁶

The family who paid $205,000 for the house near the high-scoring school bought proximity to families who earn more. The school across the boundary line, the one with the lower number, may have the teacher whose single year of instruction would add $250,000 to that family's child's lifetime earnings. The family will never know. They checked the number. The number did not measure teaching.

Sixty years

Six hundred thousand

In 1966, the United States government published the largest education study ever conducted. James Coleman and his colleagues surveyed approximately 600,000 students and 60,000 teachers across 4,000 public schools. The study was commissioned under Section 402 of the Civil Rights Act of 1964. Congress expected it to document funding inequality between schools, building the case for federal equalization of school spending.²⁷

Coleman found the opposite. Family background and socioeconomic status explained more variance in student achievement than school resources, teacher quality, or per-pupil spending. The inputs flowing from home mattered more than the inputs flowing from the district. School spending, class sizes, and physical facilities showed small effects. What the family carried through the door was the dominant variable.

The finding directly contradicted the study's purpose. The report was released on July 2, 1966, the Saturday before the Fourth of July weekend. The Johnson administration limited media coverage to findings about racial segregation.²⁷

Christopher Jencks and his colleagues reanalyzed the data in 1972 and reached the same conclusion. A team of researchers in 2010, applying modern multilevel statistical methods to the original data, added a finding that made it worse. The social class composition of a student's school, not the student's individual social class, was more than 1.75 times more important for predicting achievement. The peer effect is the mechanism. Who else is in the building matters more than what the building provides. The rating captures the composition. The rating sorts by the composition. The composition confirms the rating.²⁸

At the fiftieth anniversary, an education economist calculated the pace of progress since Coleman. At the current rate of improvement, closing the Black-white math achievement gap will take two and a half centuries.²⁷

The government has known for sixty years that school ratings primarily measure family composition. The rating system continued because the rating serves a function more powerful than any finding: it gives the sorting a name that sounds like quality.

The people who design education policy, the administrators who oversee the rating systems, the politicians who cite the numbers, do not send their children to the schools the numbers condemn. The people who know what works inside a classroom, the teachers producing two years of growth in buildings nobody chose, have no power over the rating system. The knowledge sits in one place. The power sits in another. The consequences sit in a third.²⁹

Eleven

In 1944, Britain created a system that sorted children at age eleven. Pass the exam: grammar school, the academic track, the university pathway. Fail: secondary modern, the vocational track. The exam was designed to measure intelligence and aptitude.

The Crowther Report in 1959 found that only 10 percent of children from the poorest families attended grammar school. Those who received private tutoring passed at a rate of 70 percent. Those without: 14 percent. Today, grammar schools in England enroll 6.7 percent of pupils eligible for free school meals. Non-selective schools in the same areas: 28.4 percent. The most deprived families have a six percent chance of grammar school attendance. The wealthiest one percent: 80 percent.³⁰

A test at age eleven. It was called intelligence. It measured who could afford a tutor. One hundred sixty-three grammar schools survive in England today, holding 6.7 percent free school meal students in areas where non-selective schools hold 28.4 percent. The sorting persists. It has been eighty-two years. The mechanism has not changed. It has been given a website.³⁰

The oldest standardized test in history ran for thirteen centuries on the same premise. The Chinese imperial examination, from 605 to 1905 AD. Open to any man in the empire. Merit, measured by an exam. In its first decades, 75 percent of successful candidates came from families with no prior exam holders. By its final century, the figure had fallen to 10 percent. The exam did not change. The families who could afford years of preparation, private tutors, and time away from labor gradually captured it.³¹

Three sorting systems. Eighty-two years. Thirteen hundred years. Sixty years. Each one began with the premise that the number measures quality. Each one ended measuring who could afford to produce the number.

The removal

Finland does not publish school ratings. Schools are funded by the national government through a needs-based formula that accounts for student population, special needs, and regional challenges. Local property taxes do not enter the equation. Teachers are drawn from the top 10 percent of graduates, selected through multi-stage admissions, and required to hold a master's degree.³²

The result. In PISA assessments, just seven percent of the total variation in reading performance occurs between schools. Ninety-three percent is within schools. Finnish parents choose between schools that perform within a narrow band because funding, teacher quality, and resources distribute across schools rather than concentrating where property values are highest.³³

Finland is small. Five and a half million people. More culturally homogeneous than the United States. More economically equal, with a Gini coefficient of 0.285 against America's 0.41. Its PISA scores have declined significantly since their peak in the 2000s. Math performance dropped 60 points between 2003 and 2022. The system is not a finished answer.³⁴

But the structural finding persists. Finland removed the rating. Equalized the funding. Distributed the talent. Seven percent between-school variation in a country that never published the number. The parent in Helsinki does not check a rating because there is no rating to check. The school two blocks away and the school five kilometers away perform within a narrow band because both have equally qualified teachers and equal funding. The decision was made before the parent made a choice.

The measurement added sorting.

The boundary line

The family bought the house on the right side of the line. They checked the number. They paid the premium. They moved for the schools.

The number measured their neighbors' income. The school across the boundary, the one with the lower number, may have teachers producing four times the growth per student. The family will never know. The number told them everything they needed to feel.

The parent does not say "I moved to be near wealthy people." The parent says "I moved for the schools." The rating translates economic sorting into the language of care. The feeling of making a good choice for your child is real. The number produced it. The number was not measuring what they thought.³⁵

The rating serves a real function. It makes a complex system navigable. A school is hundreds of variables. Individual teachers, classroom culture, student dynamics, leadership, programs that work and programs that fill time. The rating compresses all of it into a single digit on a screen. The compression is what makes the decision possible. The compression is also what destroys the knowledge that would tell you which school is actually teaching better. The parent who moves for the rating has been given a number. They have lost the thing the number cannot capture.³⁶

A test in England that explained less than one percent of what it claimed to measure. A cubicle in Beijing purchased for a school registration. Twenty billion dollars a year spent on tutoring in South Korea. Sixty years of data showing the same finding. Thirteen centuries of the same exam narrowing toward the same families. The rating, the score, the exam, the number on the screen. Each one calculated correctly. Each one measuring something other than what the person looking at it believed.

The rating sits on the screen. The real estate agent cited it. The property tax funded the school that produced it. Everyone looked at the number.

Nobody asked what the number was counting.