Abstract
On May 22, 2013, the Chicago Board of Education voted to close 49 elementary schools in one action. No school district in the United States had ever closed so many at once. This paper is a synthesis of the published record on that decision paired with our own descriptive analysis of public data. We did not run an experiment, interview anyone, scrape anything, or estimate a causal effect. What we did was build a map. We took the closed-school roster, 53 buildings transcribed from a primary-source list and geocoded to Chicago community areas, and set it against the 661-school system that was still operating three years later, in school year 2016-17. The picture is plain. The closures touched only 27 of the city's 77 community areas, and 49 of the 53 buildings sat on the South or West Side. Five community areas absorbed nearly two in five closures. Almost every building was a neighborhood elementary school, and not one high school was on the list. In the neighborhoods that lost the most, the schools left standing were still overwhelmingly Black and overwhelmingly poor in 2016-17. We connect those patterns to peer-reviewed work that documents who was displaced, what the action cost teachers and children, and why the under-enrollment metric that justified the closures had decades of housing segregation already baked into it. We are strict about the seam between what our map can show and what it cannot. A snapshot describes a footprint. It does not prove a cause. Every claim about harm stays where it belongs, with the University of Chicago Consortium on School Research and the published studies.
A Vote, Then a Map
The board met downtown on a Wednesday and voted. By the end of the meeting on May 22, 2013, Chicago Public Schools had approved the closing of 49 elementary schools, the largest single-year closure any American district had ever carried out [10]. The contemporaneous record is blunt about whom the decision reached. Roughly 88 percent of the affected students were African American, against 43 percent district-wide [10]. That gap was not discovered in some later study. It was visible the day of the vote, sitting in the public reporting alongside the roll call.
The headline number itself needs handling before this paper goes any further, because the count is genuinely slippery and it is better to be honest about that at the door than to have a reader trip over it halfway down the page. The slate that went into the May 22 vote was larger than the 49 that made the headlines. The action is most often summarized as 50 schools, the 49 elementary buildings plus the Mason high-school program, which was a program closed inside a building that stayed open rather than a building shut outright [11]. A primary-source list published by CBS Chicago names 53 schools and phase-outs with their street addresses, and that list is the one this paper works from, transcribed verbatim and checked name for name against other reporting from the same weeks [11]. Four elementary schools were spared at the last minute, which is part of why the 53-building roster and the 49-or-50 headline diverge. None of the geographic findings here turn on whether the right number is 49, 50, or 53. We report against the full 53-building roster throughout, and we label it that way wherever it appears, because the roster we mapped is the one with addresses we could actually place on the ground.
So set down the unit of analysis first, plainly. It is a building with a street address, geocoded to a point, assigned to a community area. When this paper says 53, it means 53 places in the city, two of which share a wall. We are not counting program closures we cannot locate. We are not relitigating the board's own arithmetic. We are mapping the buildings the public record handed us.
The second thing to set down is what kind of document this is, because that governs everything. This is not breaking news. The vote is more than a decade old, the buildings emptied out over the summer of 2013, and the scholarly record on what came after is now substantial and, in places, settled. Neither is this an outcome study. We did not follow a single child. We did not estimate what the closures did to test scores, to attendance, to graduation rates, or to the price of the houses on the blocks where the schools stood. Other people did that work, with the right tools and the right data, and this paper leans on them and names them.
What this paper is, instead, is a map and a reading of the literature that explains the territory. The honest framing fits in a paragraph, and we hold to it on every page. Every empirical number we report from our own analysis describes location. Where the closed buildings sat. How many fell in each community area. What level of school they were. How far the nearest surviving school of the same level was. What the schools still standing in those neighborhoods looked like, by race and income, in 2016-17. Every causal claim, every statement about harm to a student or a teacher, every estimate of how many children the action touched, belongs to a peer-reviewed source or to the Consortium, and each one is attributed in the text. Where our descriptive map and the published findings point the same way, we say so. Where our data simply cannot answer a question, we say that too, and we leave the question with the people who answered it properly.
That discipline matters more here than it would almost anywhere else, because the 2013 closures sit inside a live and unfinished argument about race, disinvestment, and who gets to decide a neighborhood's future. It would be easy, and dishonest, to let a clean map carry the weight of a causal claim. A map of closed buildings that all happen to land in Black neighborhoods looks like an indictment, and the peer-reviewed literature does support a strong reading of racial selectivity in which schools the city chose to close [6]. But our map by itself proves only location. It shows where the buildings were.
Be concrete about the difference, because the line between description and inference is the whole methodological spine of this paper. We can say, from our data, that 49 of the 53 closed buildings sat on the South or West Side. That is a count of points against a boundary file, and it is either right or wrong as a matter of arithmetic and geocoding, which is why we spend a whole section on how we got it right. We cannot say, from our data, that the city closed those schools because they were in Black neighborhoods. That is a causal claim about the district's decision process, and our points on a map contain no information about why any particular school was chosen. We cannot say, from our data, that the children who attended those schools were worse off afterward. That is a claim about outcomes, and we tracked no child. The step from "the closures clustered in Black, disinvested neighborhoods" to "the closures were racially selective" runs through Weber and colleagues' model [6]. The step from there to "the closures harmed the children and teachers" runs through the Consortium and the labor study [3][5]. Both steps are well supported in the literature. Neither runs through our geocoding, and we keep those seams visible instead of plastering over them.
One more thing is owed up front about the surviving-school file, because the obvious assumption about it is wrong and the whole analysis depends on not making it. The school-universe data shipped with this project is a snapshot of 661 active Chicago Public Schools from school year 2016-17 [1]. That is the post-closure system. It does not contain the 53 closed schools, which is exactly what you would expect of a roster of schools that were open in 2016, three years after the closures took effect. The closed buildings are missing from it by definition. That has a useful consequence and a hard limit. The useful consequence is that the surviving file gives us a clean comparison universe, the schools still open, where they sat, who attended them, against which the closure footprint can be read. The hard limit is that a single year cannot show a neighborhood before and after. It is one still frame taken well after the event, not a film of the change. We treat it as post-closure context and nothing more, and we flag that every time the snapshot does any work.
The decision was made on a particular date, by a particular body, against a particular list of buildings on particular streets. The rest of this paper traces the shape that decision left on the city. We are not the first to look. We are careful to say what others found before we say what we found.
A Map Drawn Decades Earlier
A closure map is never only a closure map. To see why the buildings on the 2013 roster sat where they sat, you have to pull the frame back past the one board meeting and look at the city underneath it. The work of looking has already been done, carefully and at length, by people who studied these neighborhoods directly. The numbers in this section are theirs, attributed, and they set the context that our own descriptive map then fills in.
Start with how the South and West Side Black neighborhoods came to exist at all. Between roughly 1916 and 1970, in the Great Migration, some six million Black Americans left the rural South, and several hundred thousand of them came to Chicago, fleeing the terror and the sharecropping economy of the Jim Crow South and chasing industrial work [13]. They did not get to settle wherever the jobs were. They were funneled into a narrow strip of the South Side that became known as the Black Belt, and later into a second concentration on the West Side, and they were kept there. That confinement is the first cause of everything downstream. The neighborhoods that show up at the top of the 2013 closure roster, Bronzeville and Englewood on the South Side, the Garfield Parks and Lawndale on the West, are the neighborhoods the migration was channeled into and then walled inside.
The walls were not metaphorical. In 1940 the Home Owners' Loan Corporation graded Chicago block by block for mortgage risk, and it marked the Black Belt in red [14]. Red meant hazardous, and hazardous told lenders not to write loans there. Those grades, layered on top of restrictive covenants that white property owners signed to keep Black families out, did not describe where Black Chicagoans already lived so much as fix it, by making capital flow around the red ink for a generation [15]. When the Supreme Court ruled the covenants unenforceable in 1948 and Black families began crossing the old lines anyway, they often met crowds [15]. The historian Arnold Hirsch documented the white riots that met integration at the Airport Homes in 1946 and the Fernwood Park Homes in 1947, violence that ran for nights at a stretch [12]. When a Black family moved into an apartment in suburban Cicero in 1951, a mob of thousands ransacked the building over several nights while the police stood back, and it took the National Guard to restore order [12]. The geography that produced the 2013 closure map was first drawn by appraisers and then enforced, where the maps were not enough, by crowds. These dates and events are not numbers from our own analysis; each is drawn from the named historical sources. They matter here because the structural literature on the closures rests on exactly this foundation, and a reader cannot weigh that literature, or understand why a closure map and a segregation map coincide, without it.
The anchor text on the closures themselves is Eve Ewing's ethnography of the Bronzeville closings, which argues that the 2013 action on the South Side grew out of structural racism and a long arc of housing disinvestment rather than neutral budget policy [2]. Ewing's account matters for this paper because it refuses the framing the closures were officially given. The schools were closed, in the district's telling, because they were under-enrolled, and under-enrollment was offered as a neutral fact, a building with too many empty seats to justify its cost. Ewing's contribution is to show that the emptiness was not neutral at all. The seats emptied because the people left, and the people left because of decades of policy that drained Black neighborhoods of population, of investment, and of stability. A classroom standing half-empty in Bronzeville in 2013 was reading out the cumulative result of redlining, public-housing demolition, and disinvestment that long predated any enrollment count. To close the school on the strength of that count, Ewing argues, is to treat a symptom of structural racism as the school's own failure, and to land the consequences on the very community that absorbed the original harm. She works this out at the level of one neighborhood, with the testimony of the people who lived it, and the force of the book is that the small scale does not narrow the claim. It sharpens it.
Bronzeville rewards a closer look, because the abstraction of a community-area name can flatten what is at stake. In the 1940s this stretch of the South Side held the offices of the Chicago Defender, the Regal Theater, the Provident Hospital, and the headquarters of Black-owned insurance and banking, all packed into a few square miles. It was dense because the color line gave Black Chicagoans almost nowhere else to go. The same crowding that produced that concentration of institutions also made the neighborhood a target for a particular kind of exploitation. When Black families finally pushed west and south past the old boundaries in the 1950s, many bought their homes on contract rather than with a mortgage, because the mortgage market was closed to them. The historian Beryl Satter reconstructed how that worked on the West Side [16]. A speculator bought a building cheap from a fleeing white owner, then sold it on contract to a Black family at a steep markup, holding the deed until the final payment so that one missed month could mean eviction and the loss of everything paid in [16]. The family paid like owners and were protected like nobody. Money that in a white neighborhood would have built equity in a house instead drained into a speculator's pocket, which is one of the concrete mechanisms by which the same paycheck bought a Black family less, generation over generation, in the same blocks that would later post the low enrollment counts.
Then came the towers and their demolition. The Chicago Housing Authority stacked the Robert Taylor Homes and the Stateway Gardens in a wall of high-rises along State Street, the Taylor Homes alone holding more than 11,000 people at their peak, the largest public housing development in the country [17]. The placement was not accidental. The high-rises went up along a narrow corridor hemmed by an expressway on one side and rail yards on the other, sealing the residents off from the white neighborhoods a few blocks west. Beginning in 1999 the city's Plan for Transformation reversed course and tore those towers down, calling for the demolition of some 18,000 units and promising every leaseholder a right to return [17]. Tens of thousands of residents were displaced, and far fewer came back than left, with many losing their place on the waitlist entirely [17]. By the time the school-closure list was drawn in 2013, the population that had once filled the local elementary schools had been moved out, in waves, by one public policy after another. The under-enrollment was real. It was also manufactured, slowly, by the same government that then cited it. That is Ewing's point, grounded in one place, and the quantitative literature shows the one place is the whole city.
That argument generalizes past Chicago, and the national review literature supplies the wider frame. Tieken and Auldridge-Reveles, surveying the research on school closures across the United States, reframe closure as a problem of spatial injustice, and they document that closures concentrate in Black and low-income neighborhoods as a matter of pattern rather than accident [7]. The word pattern is carrying real weight in their account. Any single closure can be explained as a local, technical decision about one under-enrolled building. The regularity only becomes undeniable when you pull back across many districts and many years, and the regularity is racial and economic. Closures cluster where Black and poor families live.
The spatial-injustice frame names the consequence, and it explains why a closure is felt as more than the sum of its enrollment math. A school is not only a place where children are taught. It is a public institution bolted to a particular spot. It holds jobs, often for people who live nearby. It is where a parent votes, where a block meets, where a child who has nothing else steady has somewhere to be at eight in the morning. When that institution is pulled out of a neighborhood, the neighborhood loses a piece of public infrastructure that is not pulled out of whiter, richer places at anything close to the same rate. The review literature's contribution is to insist that this loss be measured spatially, as a withdrawal that lands on specific ground, rather than tallied school by school as if each were a free-floating budget line [7]. The injustice, in their account, is not only that schools close. It is where they close, and where they do not. Ewing makes the same point from the inside, by showing a community that experienced a closing as a death and a grief rather than a spreadsheet adjustment, looking at the same address the district saw and not seeing the same thing at all [2].
For Chicago specifically, the question of which schools closed has been put to a quantitative model, and the answer is hard to wave off. Weber, Farmer, and Donoghue modeled the schools that closed across the city between 2000 and 2013 and found that the risk of closure rose with under-enrollment, with the Black share of a school's enrollment, and with neighborhood characteristics [6]. The finding lands hardest stated precisely, because it is the empirical hinge of the entire structural argument. Even after accounting for under-enrollment, the Black share of a school's students independently predicted whether it closed. A school was more likely to be shut if more of its children were Black, holding the official justification constant. That is selectivity, demonstrated with the right tool over the right span of years, and it is the literature that demonstrates it, not our map.
The mechanism the literature keeps returning to is the under-enrollment metric itself, and how that metric works repays spelling out, because so much rides on it. Under-enrollment is a ratio, students enrolled divided by a building's rated capacity. It looks objective. It looks like the sort of number that could not possibly carry a racial charge, since it is only bodies over seats. Picture two grade schools built in the 1960s for 600 children each. One sits in a neighborhood whose population held steady, so it still enrolls 550 and reads as nearly full. The other sits in a neighborhood that lost half its residents to demolished public housing and decades of disinvestment, so it now enrolls 250 and reads as badly under-enrolled. The ratio treats the second school as the problem. But the second school did not shrink because it failed. It shrank because the neighborhood around it was emptied out, by policy, over thirty years. The capacity in the denominator was set when the building went up. The enrollment in the numerator was set by everything that happened to the neighborhood since, which is the part the structural argument, running through Ewing, through Weber and colleagues, and through the spatial-injustice review, insists had everything to do with race [2][6][7]. The enrolled count was low because the neighborhood had lost population, and the neighborhood had lost population because of segregated housing, demolished public housing, and targeted disinvestment. By the time the district divided students by seats in 2013, the metric had already swallowed all of that history. A neutral-looking capacity ratio applied to neighborhoods with radically different pasts does not return a neutral result. It launders decades of segregation into a single figure that can be read aloud at a board meeting as a budget fact. That is the literature's claim. It is the reason a closure map and a segregation map turn out, on inspection, to be the same map.
The reporting record stretches the pattern across a longer horizon than the one 2013 action, and the figures are stark on their own. WBEZ's accounting of Chicago closings and turnarounds found that of 70,160 students in closed or turned-around schools between 2002 and 2019, 87.5 percent were Black, 61,420 students, against 533 white students across the entire seventeen-year span, with nearly every repeatedly affected building sitting in a historically African American community [9]. More than 61,000 of the affected students were Black across those seventeen years; 533 were white. A race-neutral process does not produce that distribution in a city the size of Chicago, segregated the way Chicago is segregated.
The WBEZ project also captures something our single-year snapshot cannot, which is repetition. School closing in Chicago was not a one-time event in 2013. It was a recurring instrument of policy, applied across nearly two decades, and applied disproportionately to the same kind of building in the same kind of neighborhood each time. Some blocks lost a neighborhood school, received a charter or a turnaround in its place, and then lost that too when the replacement faltered or the enrollment kept falling. A child could pass through more than one disruption before finishing grade school. That recurrence is invisible in a footprint of a single year, which is one of the honest limits of what this paper maps. We are tracing the largest single pulse, the action of May 22, 2013, and the largest pulse is worth tracing precisely. But it sat inside a longer rhythm of withdrawal that kept landing on the same ground, and the rhythm, not the pulse alone, is what the reporting record documents and what our one-year roster cannot show. We hold our claims to the pulse and credit the rhythm to WBEZ.
What "the same ground" means in practice is concrete. These are neighborhoods where, within living memory, public-housing high-rises stood and came down, where a census tract's population could fall by half across two decades, where a commercial strip that once anchored a Black middle-class community thinned into vacant lots and currency exchanges. The school in such a place is frequently among the last large public institutions still operating on its block. It still pulls children in every morning. It still employs adults from the area. It still works as a polling place, a cooling center in July, a warming center in January, the address everyone on the block can name. The literature's argument is that the under-enrollment used to justify closing that school was produced by the same long withdrawal that hollowed out everything around it, so that closing the school completes a process rather than responding neutrally to one. Ewing makes this concrete at the scale of a single community. Weber and colleagues and the review literature show the Bronzeville story is the citywide story and the national story at once [2][6][7].
We want to be modest about where our own contribution sits in all of this, because the literature is strong and we do not improve on it. The racial selectivity of the closures is established. The structural account of under-enrollment is established. The spatial-injustice frame is established. The disproportion in who was affected, over one year and across two decades, has been quantified by sources far better positioned than we are [6][7][9]. None of that is ours, and we do not claim it. What we add is narrower and still, we think, worth having. We offer a clean, reproducible, descriptive map of the 2013 footprint specifically, the 53 buildings placed on the ground and tabulated by community area, set against the surviving system three years on. The literature tells you the closures were racially and spatially selective and explains why. Our map shows you, building by building and area by area, what that selectivity looked like as a shape on the city, and what was left standing where the buildings came down. The map earns its place not by re-proving the argument the literature already proved, but by drawing that argument's footprint precisely enough to be checked against the city itself.
How We Drew the Footprint
A map is only as trustworthy as the honesty of its legend. What follows, in plain language, is what we built the footprint from, what we corrected by hand, and where the numbers stop being reliable. A careful reader uses this part to decide whether to trust the rest, and that decision should rest on a clear account of construction.
There are two real inputs, different in kind. One is the closed-school roster, 53 buildings transcribed from the CBS Chicago primary-source list, which published the full set of closed and phased-out schools with street addresses [11]. We took the addresses as given from the primary source and corroborated the roster name for name against other reporting and consolidation lists from the same period, so that the 53 we mapped is the 53 the public record names, not a set we assembled selectively. Each building on the roster is a street address. Each address is a thing we placed on the map. Overton at 221 East 49th Street. Dumas at 6650 South Ellis. Near North at 739 North Ada. The other input is the surviving-school file, a snapshot of 661 active Chicago Public Schools from school year 2016-17 shipped with this project, carrying for each school its latitude and longitude, total enrollment, low-income enrollment, enrollment by race, school type, and rating [1]. The two inputs do not overlap, and why they do not overlap is the single most important thing to grasp about the data.
The correction has to be stated flatly, because the natural assumption is wrong and the analysis collapses if it goes uncorrected. The shipped CSV is the post-closure surviving snapshot. It is not the closure map. It holds schools that were open in 2016-17, which by definition excludes schools that closed in 2013. We checked this directly against the roster. Dumas, Overton, Trumbull, Peabody, Von Humboldt, Near North, all absent from the surviving file, exactly as a list of schools open three years later should be. The file therefore cannot, on its own, show the geography of the closures, because the closed schools are not in it. What it can do, and does well, is serve as the comparison universe. It tells us what survived, where it sat, and who attended it. So the two inputs play complementary parts. The closed roster is the closure map. The surviving file is the backdrop the closure map is read against. Conflating them would be a serious error, and naming the distinction up front is how we keep from making it downstream.
Placing the closed roster on the map meant geocoding 53 street addresses, and here we owe a precise account, because geocoding is exactly the step where small errors hide and then propagate into a community-area count. Of the 53 addresses, 51 matched cleanly through the U.S. Census Bureau's public geocoder, which returns a latitude and longitude for a street address. Two did not, and we corrected both by hand with OpenStreetMap's Nominatim geocoder, flagging the fix in the matched-address field rather than burying it. One was Garfield Park Prep, which the Census file could not match at all. The school operated in the Faraday building at 3250 West Monroe, which sits in East Garfield Park, and we placed it there. The other correction is the instructive one, and it earns the full telling, because it is the kind of error that, left in, would have moved a school to the wrong side of the city. Trumbull's address is 5200 North Ashland. The Census geocoder snapped it to 5200 South, flipping a far-North-Side school onto the South Side. Trumbull is genuinely in Edgewater, on the far North Side, and we corrected it back there. That one fix is the reason Trumbull turns up later as the lone far-North closure instead of disappearing into a South Side tally. It is a small demonstration of why a directional audit was not optional.
Community areas were not eyeballed, and the reason for the formality matters. Chicago's 77 community areas are a fixed, official geography, long stable, which makes them the natural unit for talking about where something happened in the city. But assigning an address to one by eye, or by ZIP code, invites exactly the kind of small error that accumulates into a wrong headline. So we assigned each geocoded point to one of the 77 areas by a point-in-polygon test against the city's official boundary file, the standard authority for this kind of assignment. A point either falls inside a given community-area polygon or it does not, and each point falls inside exactly one, which removes the judgment call. The polygon a point lands in is the area we recorded, full stop. To guard against the precise failure the Trumbull case exposed, where a geocoder's mistake about north versus south would have moved a school to the wrong side of the city while still returning plausible-looking coordinates, we ran a directional-flip audit across all 53 assignments, checking for any address whose north-south or east-west designation looked inconsistent with the area it landed in. After the two Nominatim corrections, the audit found no remaining directional errors. We mention it not to claim the map is flawless but to be specific about the one failure mode we hunted for and cleared, because a concentration figure as lopsided as 49 of 53 is exactly the kind of result a hidden directional error could manufacture or inflate, and a reader is entitled to know we looked.
Two pairs of closed schools share a single building address, and we flag this as genuine co-location rather than let it read as a data error or a double count. Roque De Duprey and Von Humboldt were both at 2620 West Hirsch. Williams Middle Prep and Williams Multiplex were both at 2710 South Dearborn. In each case, two distinct schools operated out of one building, a real arrangement in Chicago Public Schools, where a single facility could house, for instance, a neighborhood elementary program and a separate middle-grade or specialty program under different names and different leadership. That is not an artifact of our transcription. It is how the system actually organized those buildings. We carry both schools in the roster because both were on the closure list, and we note the shared address so that anyone re-deriving the geography understands why two of our points sit exactly on top of each other. Two closures at one location is not one closure counted twice. It is two programs, each of which served children and employed teachers, both ended in the same action, and collapsing them into a single point would erase one of the two communities that lost its school. We keep both, and we keep the shared coordinate honest by naming it.
This is also where the divergence between 53 and the headline counts becomes legible rather than confusing. The board's action is most often summarized as 49 or 50 schools, and the roster we mapped holds 53 buildings, and both can be right at once because they are counting slightly different things. The headline counts reflect the schools as administrative units after the last-minute reprieves, with a program closure inside a surviving building handled one way and co-located programs handled another. Our roster counts the buildings and programs with street addresses that the primary-source list named, transcribed before any of that administrative tidying. We do not claim our 53 is the one true number. We claim it is the roster the public record gave us with addresses we could place, and that none of our geographic findings shift whether a reader prefers 49, 50, or 53, because the concentration on the South and West Sides, the clustering in a few community areas, and the all-elementary character of the action hold under every one of those counts.
On the surviving side we made one exclusion, small but named. One active record in the 661-school file, the YCCS-Virtual high school, carries zero enrollment. A school with no students cannot meaningfully enter an enrollment-weighted demographic calculation or a majority-Black determination, since those quantities are undefined or distorted at zero. We dropped that single record from all rate calculations, which is why the surviving-system rates in this paper run over 660 of 661 schools rather than the full 661. The count of 661 is the right number for active schools in the snapshot. The count of 660 is the right denominator for schools that actually enrolled students. We use each where it belongs.
Apart from that one record, the surviving file was clean in the fields we needed, which is part of trustworthiness and something a reader cannot check from the outside. Every active school in the snapshot carried a community-area assignment, a total enrollment, a low-income enrollment, and a racial breakdown, with no missing values in those columns. We did not impute anything, drop anything for incompleteness beyond the single zero-enrollment record, or fill a gap with an assumption. The demographic shares we report are computed directly from the enrollment counts in the file, weighted by enrollment so that a large school counts for more than a small one, which is the right weighting for a question about where students were rather than where buildings were. None of this makes the snapshot more than it is, a single post-closure year. It does mean that within that year, the numbers are the file's own and not our reconstruction of it.
That leaves the point that governs every figure in the sections to come. Our numbers are descriptive geometry and counts. They report how many closed buildings fell in each community area, which side of the city they sat on, what level of school they were, how the surviving schools in those areas broke down by race and income, and how far each closed building was, in a straight line, from the nearest surviving school of the same level. Not one of those quantities is an outcome. None of them follows a child, measures a test score, tracks a family that moved, or estimates what the closures caused. The outcomes belong to the cited studies, and where this paper discusses harm to students or teachers it is reporting their findings under their names, not deriving anything from our file. We built a footprint. The footprint is real, it is reproducible from the named inputs, and it is exactly as much as our data can honestly carry.
Forty-Nine of Fifty-Three, on Two Sides of Town
Before any subtlety, the map shows concentration. The 53 closures touched only 27 of Chicago's 77 community areas. More than two-thirds of the city's neighborhoods saw no closure on the 2013 roster at all. Of the 53 buildings, 49 sat on the South or West Side, which is 92.5 percent, with the South Side holding 27, or 50.9 percent, and the West Side 22, or 41.5 percent [1]. The remaining four, 7.5 percent of the roster, fell on the North or Central side [1]. That is the headline, and it is a descriptive fact about where the buildings were, not yet a claim about why.
The split between the two hard-hit sides repays a second look. The South Side took the largest single share, a hair over half the roster, but the West Side, much smaller in land and population, took more than two in five closures of its own. The West Side of Chicago is, roughly, the band of community areas running west from the Loop through the Near West Side, the Garfield Parks, North and South Lawndale, Humboldt Park, and Austin out to the city limit. It is a fraction of the city's footprint, and it absorbed 22 of 53 closures. The concentration here is partly the contrast between two sides and the third, and partly how much the West Side, in particular, took relative to its size, which is a point the side-of-city pie carries better than a sentence does.
Forty-nine of the fifty-three closures fell on the South or West Side
The four exceptions deserve their names, both because they are the whole of the non-South-and-West footprint and because naming them keeps the 92.5 percent honest instead of letting it round up to all. Manierre sat in the Near North Side. Stewart and Stockton were both in Uptown. Trumbull was in Edgewater, the only far-North closure on the entire roster [1]. That last one is the school from the methods section, the one the Census geocoder tried to flip to the South Side before we set it back in Edgewater. Its appearance here as a genuine North Side outlier is the payoff of that correction. Had we let the error stand, the map would have shown an even tighter South-and-West concentration than the true 49 of 53. There is a quiet lesson in that. A clean-looking concentration figure can be made cleaner by a mistake, and the work of catching the mistake runs against the grain of the headline, which is exactly why it has to be done.
What the concentration means is the place where we have to keep the line brightest between our count and the literature's estimates, because they answer different questions and only one of them is ours. We report where the buildings were. We do not report how many children the action reached, because our file cannot tell us. The people who did answer that question reached a number that dwarfs the building count. The UIC community-mapping reanalysis by Radinsky and Waitoller found that the district had underestimated the number of affected students by 42 percent, putting the true figure above 47,000 students, roughly 90 percent of them African American and about 15 percent of them children with individualized education programs, and concluded that 27 percent of all Black children in the city were affected by the action [8]. More than a quarter of every Black child in Chicago was touched by a single year's closures. Our 53 points on a map cannot produce that number and do not try. The UIC reanalysis estimates how many children the action reached. We report where the buildings it closed were standing. Both are true. They are about the same event. They are not the same measurement, and we keep them apart on purpose.
A word is owed on the side-of-city groupings, since they carry the headline and a reader is right to ask what they are. South Side, West Side, and North or Central are standard descriptive labels in Chicago, built up from the official community areas using the conventional area-to-side correspondence. They are not analytic categories, and we do not treat a side as a unit with properties of its own. They are a coarse way of folding 27 community areas into a picture a reader can hold, which is that the closures fell overwhelmingly on the South and West Sides and barely grazed the North. We use them for that and nothing else. The finer-grained community-area counts, the real unit, come next.
For all its bluntness, the side-of-city cut hides as much as it shows. Forty-nine of 53 on the South or West Side is striking, but those two sides between them hold a large share of the city, and a closure landing somewhere in that expanse could in principle have been spread evenly across dozens of neighborhoods. It was not. Inside the 27 community areas the closures touched, the buildings clustered far more tightly than an even spread would predict, with a handful of areas taking a wildly disproportionate share.
The Tight Cluster of the Top Ten
Zoom from the side of the city to the named neighborhoods and the concentration sharpens from striking to severe. The top five community areas on the roster were West Town with 5 closures, then West Englewood, Austin, East Garfield Park, and Douglas with 4 apiece. Those five areas held 21 of the 53 closures between them, which is 39.6 percent of the entire roster [1]. Five of the city's 77 community areas absorbed nearly two in five closures. Widen the lens one notch to the top ten, adding West Garfield Park and West Pullman at 3 each and then Grand Boulevard, Woodlawn, North Lawndale, Near West Side, Auburn Gresham, and Uptown, and the top ten together held 33 of 53, which is 62.3 percent [1]. Nearly two-thirds of every building on the roster fell in ten neighborhoods.
It helps to set that against what an even spread would look like. The closures touched 27 community areas. If the 53 buildings had been distributed evenly across those 27, no area would hold more than two or three, and the top five would account for something closer to a fifth of the roster than two fifths. Instead the top five hold 39.6 percent and the top ten hold 62.3 percent, which is the signature of clustering, not spread. The closures did not rain evenly even across the neighborhoods they reached. They fell hardest on a short list within the list. That second layer of concentration, an uneven distribution inside an already narrow set of areas, is the finding the community-area cut adds to the blunter side-of-city one, and it is a fact about our own counts that holds regardless of any claim about why.
2013 closures concentrated in a few South and West Side community areas
The named places carry their histories, and those histories are the ones the structural literature rests on. Take Austin, on the far West Side, which lost 4 schools. In 1960 Austin was almost entirely white [18]. Over the following two decades it went through some of the most aggressive blockbusting in the city, as real-estate agents stoked white panic to buy low and sell high to incoming Black families, and the neighborhood flipped from white to Black faster than almost anywhere in Chicago [18]. North Lawndale, which lost 2, had been a dense Jewish community in the 1940s, then became overwhelmingly Black during the migration, and then lost a staggering share of its population and its industrial jobs after the 1968 unrest that followed the assassination of Martin Luther King Jr., who had himself moved into a North Lawndale flat in 1966 to dramatize the conditions of the slums there [18]. These are not interchangeable dots. Each name carries a specific sequence of confinement, flipping, disinvestment, and population loss, and the sequence is exactly what produced the empty seats the 2013 closures counted. Most of the top areas are the historically Black, disinvested South and West Side neighborhoods the literature describes, the Englewoods and Garfield Parks and Lawndales where decades of population loss produced the under-enrollment the closures cited [2][6]. West Englewood, East Garfield Park, West Garfield Park, North Lawndale, Grand Boulevard, Woodlawn, West Pullman, Austin, and Auburn Gresham are the places the structural account is about. Their position near the top of a closure roster is the footprint of the selectivity the peer-reviewed work establishes [6]. The one entry that does not fit that description sits at the very top of the list. West Town, with 5 closures, the most of any community area, is not a disinvested South or West Side neighborhood in the same sense. Its eastern edge, around Wicker Park and East Ukrainian Village, has been gentrifying for years, and a closure there plausibly reads off a different process than a closure in Englewood does, one with more to do with rising rents and a changing population than with the long withdrawal Ewing traces in Bronzeville.
We name that distinction, and then we stop, because our data cannot adjudicate it. A community-area closure count is a count. It does not carry the reason a school emptied, and a school in gentrifying West Town and a school in disinvested East Garfield Park show up on our map as the same kind of dot. The literature can speak to the differing mechanisms. Weber and colleagues' model is built precisely to separate the effect of under-enrollment from the effect of neighborhood and racial composition [6]. Our descriptive footprint cannot do that separation. It would be a mistake, and a tempting one, to let West Town's position at the top stand in for a single story about why these particular neighborhoods lost their schools. The honest statement is narrower. Five neighborhoods absorbed 21 closures and ten absorbed 33. Most of them fit the disinvestment pattern the literature documents, and one prominent one does not, and our count alone cannot tell you which mechanism produced any individual closure.
The raw counts have a blind spot worth correcting before we read too much into them, which is that a community area with many schools and a community area with few are not on the same footing when each loses, say, four. There is a second way to read the same concentration that gets at where the system thinned hardest relative to what was left, and it surfaces a few areas the raw counts hide. Take the ratio of closed schools to surviving schools in each community area, and you measure not the absolute number of closures but the depth of the cut against the local base. By that read, Fuller Park stands alone at the top, with 1 closure against 1 surviving school, a ratio of 1.00, the only area in the city where the closures matched the entire surviving footprint one for one [1]. West Garfield Park follows at 0.50, 3 closures against 6 survivors, one closed school for every two that stayed open [1]. A cluster of South and West Side areas sits at 0.33, among them West Englewood with 4 closures against 12 survivors, Douglas with 4 against 12, West Pullman with 3 against 9, and South Deering with 1 against 3, each losing one school for every three left standing [1]. Woodlawn, East Garfield Park, Grand Boulevard, and Uptown sit just behind at 0.29 [1]. Calumet Heights, with 1 closure against 5 survivors, sits near the bottom of this list at 0.20, a reminder that an area can appear on the closure roster at all and still have lost a relatively shallow slice of its local schools [1].
The two readings disagree in instructive places, and the disagreement is the point. West Town, with the most closures of any area at 5, carries a relatively low ratio of 0.21, because 24 schools survived there, so the absolute count makes it look like the epicenter while the ratio says the local system there was thinned less than in a dozen other places [1]. Fuller Park runs the other way. A single closure barely registers on the absolute count, yet because almost nothing survived alongside it, the ratio puts it at the very top [1]. Neither reading is the true one. They measure different things. The count answers how many buildings a neighborhood lost. The ratio answers how large that loss was against what the neighborhood had. A reader who wants the human meaning of either number has to go to the studies, because both are still only inventory.
The ratio has to be read descriptively, and the pull to over-read it is strong, so we will be exact about what it does and does not mean. A high closed-to-surviving ratio means the footprint cut deep in that area relative to its surviving base. It does not mean that any particular child there fared worse, that displaced students had farther to travel, or that the neighborhood was harmed more severely. Those are outcome questions, and outcome questions belong to the Consortium and the published studies, not to a ratio of two counts. Fuller Park having a ratio of 1.00 tells you the closure footprint there was as large as everything that survived, a fact about geometry and inventory. It tells you nothing on its own about what happened to the children who had gone to the closed school, which is a fact about people that our data does not contain. The number marks where to look. It is not itself the answer to what one would find there.
Read at the community-area scale, then, the map shows a footprint concentrated twice over, first onto two sides of the city and then onto a short list of neighborhoods inside them, and it shows where that footprint cut deepest against the local base. What it has not yet shown is what kind of school was actually closed. A roster of 53 buildings could in principle be high schools, elementary schools, specialized programs, or some mix, and the civic meaning of the action turns substantially on the answer.
[1]: Author analysis of 2013 CPS closed-school list (53 schools, with street addresses) plus repo CPS active-school snapshot (661 schools, SY2016-17) as comparison universe. Chicago Public Schools / Chicago Board of Education action (closed list, primary source via CBS Chicago and Education Week, corroborated by UIC Great Cities Institute); City of Chicago Data Portal for the CPS school universe; repo file already shipped. Reproducible files at rooted-forward.org/research/data/chicago-2013-school-closures-geography.
[2]: Ewing, Eve L. (2018). Ghosts in the Schoolyard. Racism and School Closings on Chicago's South Side. Chicago, University of Chicago Press.
[3]: Lee, Helen, and Lauren Sartain (2020). School Closures in Chicago. What Happened to the Teachers? Educational Evaluation and Policy Analysis, 42(3), 331-353. DOI 10.3102/0162373720922218.
[4]: Gwynne, Julia A., and Marisa de la Torre (2009). When Schools Close. Effects on Displaced Students in Chicago Public Schools. Chicago, UChicago Consortium on School Research.
[5]: Gordon, Molly F., Marisa de la Torre, Jennifer R. Cowhy, Paul T. Moore, Lauren Sartain, and David Knight (2018). School Closings in Chicago. Staff and Student Experiences and Academic Outcomes. Chicago, UChicago Consortium on School Research.
[6]: Weber, Rachel, Stephanie Farmer, and Mary Donoghue (2020). Predicting School Closures in an Era of Austerity. The Case of Chicago. Urban Affairs Review, 56(5), 1487-1521. DOI 10.1177/1078087418802359.
[7]: Tieken, Mara Casey, and Trevor Ray Auldridge-Reveles (2019). Rethinking the School Closure Research. School Closure as Spatial Injustice. Review of Educational Research, 89(6), 917-953. DOI 10.3102/0034654319877151.
[8]: Radinsky, Josh, and Federico R. Waitoller (UIC College of Education). Community mapping reanalysis of the 2013 CPS closures. https://education.uic.edu/news-stories/school-closure-data/
[9]: WBEZ (Sarah Karp et al.). A Generation of School Closings (interactive data project). https://interactive.wbez.org/generation-school-closings/
[10]: Maxwell, Lesli A. (2013, May 22). Chicago Board Votes to Close 49 Elementary Schools. Education Week.
[11]: CBS Chicago (2013). List of Chicago Public School Closings (53 schools with addresses). https://www.cbsnews.com/chicago/news/list-of-chicago-public-school-closings/
[12]: Hirsch, Arnold R. (1983). Making the Second Ghetto. Race and Housing in Chicago, 1940-1960. Cambridge, Cambridge University Press. Documents the white riots at the Airport Homes (1946) and Fernwood Park Homes (1947) and the Cicero riot of 1951.
[13]: Wilkerson, Isabel (2010). The Warmth of Other Suns. The Epic Story of America's Great Migration. New York, Random House. For the scale and dates of the Great Migration (roughly 1916 to 1970) and the movement of Black Southerners to Chicago.
[14]: Nelson, Robert K., LaDale Winling, et al. Mapping Inequality. Redlining in New Deal America. Digital Scholarship Lab, University of Richmond. HOLC area description and "residential security" maps of Chicago, 1940. https://dsl.richmond.edu/panorama/redlining/
[15]: Rothstein, Richard (2017). The Color of Law. A Forgotten History of How Our Government Segregated America. New York, Liveright. On HOLC grading, restrictive covenants, and Shelley v. Kraemer (1948), which ruled racially restrictive covenants judicially unenforceable.
[16]: Satter, Beryl (2009). Family Properties. Race, Real Estate, and the Exploitation of Black Urban America. New York, Metropolitan Books. On contract selling on Chicago's West Side in the 1950s and 1960s.
[17]: Chicago Housing Authority, Plan for Transformation (launched 1999), CHA program record and annual reporting. On the Robert Taylor Homes (peak population over 11,000), the planned demolition of roughly 18,000 public-housing units, and the right-to-return policy. https://www.thecha.org/about/plan-forward/plan-transformation
[18]: Local Community Fact Book, Chicago Metropolitan Area (Chicago Fact Book Consortium, University of Illinois at Chicago), community-area profiles for Austin (No. 25) and North Lawndale (No. 29), with decennial census racial composition. On Austin's racial composition in 1960 and its rapid racial transition, and on North Lawndale's population and the 1966 Chicago Freedom Movement and 1968 unrest.
Neighborhood Grade Schools, Not High Schools
The answer is unambiguous in our roster. Of the 53 buildings on the 2013 list, 50 were elementary schools, which is 94.3 percent, and not one was a high school [1]. The three remaining records are middle-grade programs, Canter, Pershing West, and Williams Middle Prep, and each of those sat paired with an elementary building rather than standing alone as a comprehensive secondary school. The action reached down into the part of the system that serves the youngest children, and it left every high school in the city untouched.
The closures were almost entirely neighborhood elementary schools
That distinction is not a technicality, and the difference is geographic before it is anything else. The youngest children walk the shortest distances. They are the least able to manage a long or unfamiliar commute on their own, and they depend most on the adults who happen to be near. A high school draws from a wide swath of the city, and its students are old enough to take a bus or a train across neighborhood lines, so closing one redistributes teenagers across a regional catchment. Closing a neighborhood elementary school does something more local and more intimate. It removes the institution that a five-year-old and a parent organized the morning around, the one within walking distance, and it pushes the youngest, least mobile children across boundaries that a teenager could absorb but a small child cannot. The geography of harm, if there is harm, is therefore concentrated tighter for an elementary closure than for a high-school closure, which is part of why the all-elementary character of the 2013 action is not a footnote but a feature of what kind of action it was.
A neighborhood elementary building also anchors the block it sits on. It is the place a parent can reach on foot in the time a lunch break allows, the corner where families who otherwise never cross paths come to recognize one another, the polling place, the after-school room, the gym that opens for the community on a Saturday. When the literature describes what closing these particular buildings did to the blocks around them, it is describing institutions of that kind, woven into a few square blocks, not the larger and more regional draw of a high school. Our data establishes the elementary character of the 2013 action and its scale. What that character meant for the people inside those buildings is documented elsewhere, and it is documented carefully.
The fullest account is the UChicago Consortium's study of the 2013 closures themselves. That work found that students from the closed schools lost, on average, about 1.5 months of learning in reading and about 2 months in math during the year the closures were announced, that the math declines were still detectable four years later, and that the transition into new schools was disorganized and stressful for the children living through it [5]. The Consortium is careful about the variation behind those averages, and so are we in repeating them. The figures are theirs, drawn from a longitudinal design our snapshot cannot reproduce, and we cite them as the established record, not as anything our own geometry could show. The phrase "about two months of math" is easy to read past. It is a national norm's worth of arithmetic, fractions, place value, the foundations the rest of school math is built on, lost in a single year by children who had no say in the decision that disrupted them.
The pattern the 2013 study documents did not come out of nowhere. An earlier Consortium analysis of Chicago closures in the 2000s, covering 18 schools and 5,445 displaced students, found that roughly 8 in 10 of those students landed in schools that ranked in the bottom half of the system on academic measures, and that the factor most associated with whether a displaced child's achievement fell was the quality of the school that received them [4]. That earlier study is the hinge on which the 2013 findings turn, because it isolates the mechanism. Moving children is not, by its evidence, automatically harmful or automatically benign. It depends heavily on where they end up. A displaced child who lands in a genuinely stronger school can gain. A displaced child who lands in a school no better than the one that closed mostly absorbs the cost of the move with none of the benefit. In the Chicago cases the receiving schools were frequently no stronger, which is the practical reason the achievement effects came out negative rather than neutral. The under-enrollment metric that selected which schools closed did nothing to guarantee that the surviving schools nearby were any better, since they sat in the same disinvested neighborhoods and drew from the same depleted enrollment. So the children were, in many cases, moved sideways. That chain of reasoning, from selection by under-enrollment to receiving-school quality to student outcome, is the Consortium's, built on a longitudinal design our snapshot cannot reproduce. Our roster can say only that the 2013 action moved a great many elementary-age children at once, and that the straight-line distances between the closed buildings and the nearest surviving schools, which we measure later, were mostly short. Whether the schools at the ends of those short distances were any better is a question our data does not touch and theirs does.
The teachers were moved too, and the labor side of the ledger has its own peer-reviewed accounting. Lee and Sartain studied what the closures did to the adults in the buildings, describing an action that covered on the order of 47 to 49 elementary schools, roughly 13,000 students, and about 900 teachers [3]. Using a difference-in-differences design, comparing closed schools against similar schools that stayed open, they found that the closures nearly doubled the rate at which teachers left their schools [3]. A closure is sometimes pictured as relocating a fixed staff intact down the street. The evidence points the other way. The disruption pushed experienced teachers out of those classrooms at roughly twice the baseline rate, with everything that implies for the children who would otherwise have learned from them, since a teacher's experience is among the few school inputs that research consistently ties to how much students learn. Again the causal claim is theirs and the design is theirs. We note only that the scale they describe, thousands of students and hundreds of teachers, is consistent with the elementary-heavy roster we mapped.
The line we hold here is the line that runs through the whole paper. Our data confirms what kind of schools were closed, that they were neighborhood grade schools, and that there were enough of them to make this the largest such action any American district has taken in a single year. It confirms the elementary character and the scale. It does not, and cannot, establish what the closures did to the children and teachers inside those buildings. For that the record points to the Consortium and to Lee and Sartain, and we let their findings stand as theirs.
What Was Left Standing
The surviving-school file describes Chicago Public Schools as it stood in school year 2016-17, three years after the closures, and the citywide profile is itself a portrait of a segregated system. Of the 660 active schools with any enrollment, 313 were majority-Black, meaning Black children made up at least half the student body, which is 47.4 percent of those schools [1]. Weighting by enrollment, the share of students across the system who were low-income was 77.8 percent, and the share who were Black was 36.8 percent [1]. Set those last two beside each other and a familiar geometry appears. Black students were a bit over a third of the system, yet they filled nearly half of its schools to majority, which is what a segregated enrollment looks like in summary. The geography of those schools tilts the same way the closures did. One ZIP code, 60623, covering North and South Lawndale on the West Side, held 36 surviving schools, more than any other ZIP in the city [1]. The system that remained was concentrated, poor, and across large stretches almost entirely Black, before we narrow the lens at all.
Narrowing it sharpens the picture hard. The ten community areas that had lost the most schools in 2013 still held a real surviving system three years later, ranging from West Garfield Park with 6 schools left and Woodlawn and Grand Boulevard with 7 each, up through West Pullman with 9, West Englewood and Douglas with 12, East Garfield Park with 14, North Lawndale with 21, and Austin and West Town with 24 apiece [1]. So these were not neighborhoods stripped bare. Schools remained, in some cases many of them. The point is what those surviving schools looked like. West Englewood stood at 86.0 percent Black, Austin at 88.8, Douglas at 93.3, North Lawndale at 94.7, East Garfield Park at 96.6, and West Garfield Park at 98.3, with West Pullman, Grand Boulevard, and Woodlawn all sitting inside that same high band at 90.7, 97.2, and 97.7 [1]. Across all ten of the hardest-hit areas, the low-income share of the surviving schools ran from 74.0 percent in West Town up to 90.9 percent in North Lawndale, with most clustered in the high 80s [1]. These are not places where a few schools closed and a more mixed set carried on. They are places where the buildings that survived serve student bodies that are, by these measures, close to homogeneous. West Garfield Park is the limit case. A surviving system that is 98.3 percent Black, drawn from the six schools left standing there, is for all practical purposes a system with almost no white or Latino or Asian children in it at all.
Schools left standing in the hardest-hit areas stayed majority-Black and low-income
The caveat has to come in the same breath as the finding, because the timing invites a mistake. This is a snapshot from school year 2016-17, three years after the closures took effect. It is post-closure context, not a before-and-after panel. It cannot show what these neighborhoods or their schools looked like in 2012, and it cannot tell us how much of the 2016-17 composition the closures produced as against how much was already there before the first building emptied. We did not assemble a longitudinal file. The one school-universe file the repo carries is this single year. So the snapshot can establish that the surviving schools in the hardest-hit areas are Black and low-income today. It cannot establish that they became so because of the closures, and we do not claim that they did. The one entry on the top-ten list that breaks the pattern says the same thing from the other direction. West Town's surviving schools were only 21.7 percent Black and 74.0 percent low-income, far off the band the other nine share [1], which is consistent with West Town being a different kind of neighborhood than Englewood, the way the closure counts already hinted, and is one more reason not to read a single causal story into the map.
What the literature offers in place of the causal claim our snapshot cannot make is the longer history. The peer-reviewed work reads the same segregated geography we are looking at as the product of decades of housing discrimination, disinvestment, and population loss, the structural arrangement Ewing traces on the South Side and the national review literature describes as the spatial pattern of American school closure more broadly [2][7]. That the schools still standing in West Garfield Park or North Lawndale are 95 or 98 percent Black is, on that reading, a feature of the city these neighborhoods sit inside, not something our snapshot can pin on the 2013 action. The snapshot documents the standing footprint. The explanation for the footprint's shape belongs to the historical and structural account the cited authors build, and we borrow none of their causal weight for our own counts.
How Far Is the Next School, and What That Cannot Tell Us
The last thing our data can describe is distance, and it is the result most likely to be over-read, so it bears stating precisely and then fencing in just as precisely. For each of the 53 closed schools we measured the straight-line distance to the nearest surviving school of the same level in 2016-17, elementary to elementary and middle to middle. The median came to 0.18 miles. The mean was 0.27. More than nine in ten of the closed schools, 90.6 percent, sat within half a mile of a surviving same-level school, and 94.3 percent sat within a mile [1]. By the crude measure of how the dots fall, most of these closures happened in parts of the city dense enough with schools that another building of the same level stood close by, often a few blocks off.
A handful of distances were much larger, and they are honest to name because they have an honest explanation. Williams Middle Prep was 2.27 miles from the nearest surviving middle school, and Pershing West was 1.67 [1]. Both are middle-grade programs, and the reason their nearest same-level neighbor sits so far off is arithmetic, not hardship. Only ten standalone middle schools survived in the entire system in 2016-17, so any measurement that restricts the match to standalone middle schools is working from a very thin field and will turn up large gaps almost by construction. Those distances are an artifact of how few middle schools there were to match against. They are not evidence that displaced children walked two miles to class, and we do not read them that way. Canter, the next farthest at 1.10 miles, is the same story, a middle-grade program with few peers to match against [1].
We chose the straight-line measure deliberately, and we chose to match each closed school only to a surviving school of the same level, an elementary to an elementary, a middle-grade program to a middle-grade program, because matching a closed grade school to the nearest surviving high school would measure nothing a displaced child experiences. Even with those choices made carefully, the more important point is what a straight-line distance is not. It is geometry between two pins, and a city is not geometry. The line says nothing about the actual walking route a child takes, which bends around blocks and rail embankments and the river and is always longer than the straight line, sometimes much longer where a viaduct or an interstate forces a detour. It says nothing about which school a displaced child was actually assigned to, since the building nearest on a map is not necessarily the designated welcoming school the district named, and the assignment, not the nearest pin, is what determined where a child actually went. It says nothing about whether the path between two points crosses a gang or street boundary that makes the short route the unsafe one, which was the exact concern the district's Safe Passage program was built around, hiring adults to stand watch along marked routes precisely because the shortest line between an old school and a new one sometimes ran through contested ground. And it says nothing at all about any outcome for any student. We report the distance because it is part of an honest description of the footprint, and the median of 0.18 miles is a genuine fact about how densely the closures fell. We fence it because the same number, read carelessly, would seem to say the closures were therefore low-cost, and that is not a claim the distance can support. The data caveats name each of these limits, and each one is real. Whether the closures lengthened children's commutes in practice, routed them across dangerous ground, or moved them to schools no better than the ones they left is a set of questions the Consortium and the peer-reviewed studies took up directly, and the answers are theirs, not ours [4][5].
It helps to gather the governing limits in one place, since they apply to everything above. There is the count nuance, that the roster holds 53 buildings while the action is often summarized as 50 or 49 schools depending on the last-minute reprieves and on how a program closure inside a surviving building is counted, and that no geographic finding here depends on which headline number a reader prefers, because we report against the full 53 throughout and label it so. There is the timing, that the surviving-school file is a 2016-17 snapshot three years after the closures and is therefore post-closure context, not a panel that could support a before-and-after comparison. And there is the nature of every number we have reported, that it is descriptive geometry and counts, where buildings stood, how many fell in each place, what kind of schools they were, how far apart the dots sit. We ran no experiment. We built no model. We estimated no effect. On cause and consequence we defer entirely to the published record, and where our description and their findings point the same way we treat that as corroboration of a pattern, not as proof we produced. Claiming less is the whole point of an honest map.
The Shape the Decision Left
The descriptive bottom line is small enough to hold in a sentence. On a single day in 2013 the Chicago Board of Education closed a set of schools that were almost entirely neighborhood grade schools, concentrated overwhelmingly in a short list of historically Black, disinvested community areas on the South and West Sides, and the system left standing in those same areas is still, three years on, overwhelmingly Black and low-income. That is what our roster and the surviving-school snapshot show when they are mapped against each other. It is all they show.
The rest of what this paper carries is borrowed and attributed. The harms to the students who were displaced and to the teachers who left come from the Consortium and from the peer-reviewed labor study, not from anything we measured [3][5]. The estimate that the action reached far more children than the district acknowledged, and that roughly a quarter of all Black children in the city were affected, comes from the UIC community-mapping reanalysis, not from us [8]. The reading that the under-enrollment driving the closures was itself the residue of decades of segregation comes from Ewing and from Weber and colleagues and from the spatial-injustice review, not from our counts [2][6][7]. We have been deliberate about keeping that line bright in every section, because the temptation in this kind of work is to let a vivid map stand in for a verdict on cause, and a map cannot bear that load.
Still, the geography matters on its own terms, and the reason is one the literature states more fully than we will. Where a city chooses to close its schools is a record of where it has already pulled back. The under-enrollment that put a building on the list was, in the peer-reviewed account, the downstream trace of decades of housing segregation and population loss on the same blocks, so the closure map reads as a second draft of an older map of disinvestment [2][7]. We are not restating that conclusion as ours. We are pointing at the plain fact that the 2013 footprint lands exactly where that older geography would predict, and leaving the explanation to the authors who built it. The map is consistent with the history. It does not replace the history's argument, and it was never meant to.
There is also the matter of what a closure leaves physically behind, which our snapshot is silent on but which belongs in any honest reckoning of the geography. A closed school is more than a service withdrawn. It is a large municipal building, often most of a city block, that has to be sold, reused, or left to sit. Many of the 2013 buildings sat empty for years, hard to sell precisely because they stood in the disinvested neighborhoods the closures had targeted, where there was little market demand for a former school. An empty school is its own kind of disinvestment, a dark anchor on a block that already had too few. The spatial-injustice frame applies here as squarely as it does to the closing itself, since the burden of a vacant building, like the burden of the closing, fell on the South and West Sides and not on the North [7]. Our data cannot speak to the afterlives of these 53 buildings, and we do not pretend it can. We note only that the footprint we mapped did not vanish when the schools did. It became a different shape on the same ground.
The files behind every number here are public, and they are meant to be checked. The closed-school roster, the surviving-school snapshot, the geocoding, the community-area assignments, and the counts that follow from them are all posted, so that another reader can redraw the same map, test the same distances, and decide whether the description holds [1]. We would rather the map be verified than believed.
So it comes back to the day it started. On May 22, 2013, a board vote drew a line through 53 buildings, nearly all of them grade schools, nearly all of them on two sides of one city, and the streets that line traced run from West Town to West Englewood, through Garfield Park and Lawndale, through Douglas and Austin, the same blocks where the schools that remain still teach children who are almost all Black and almost all poor. That is the shape the decision left on the map. What it left in the lives of the people it moved is written in the studies, and it is not ours to claim.
Citations
18 sources cited.
Primary Sources
- 1.Author analysis of 2013 CPS closed-school list (53 schools, with street addresses) + repo CPS active-school snapshot (661 schools, SY2016-17) as comparison universe. Chicago Public Schools / Chicago Board of Education action (closed list, primary source via CBS Chicago and Education Week, corroborated by UIC Great Cities Institute); City of Chicago Data Portal for the CPS school universe; repo file already shipped. Reproducible files at rooted-forward.org/research/data/chicago-2013-school-closures-geography.
Secondary Sources
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
- 12.
- 13.
- 14.
- 15.
- 16.
- 17.
- 18.