Forty Years of Public Subsidy on San Francisco's Western Addition Renewal Footprint

Abstract

The blocks the San Francisco Redevelopment Agency once labeled Western Addition Area 2 were emptied on purpose. Between the early 1950s and 1970, an agency working hand in glove with downtown business cleared a majority-Black and Japanese American district, and by the city's own later reckoning roughly 5,893 mostly Black households were removed ^[10]. This paper does two things that do not touch. It gathers the published record of what postwar renewal did to that neighborhood and to neighborhoods like it across the country, and it offers a descriptive read of one real public dataset, the San Francisco MOHCD and OCII Affordable Housing Portfolio drawn from DataSF, to describe what public subsidy has since rebuilt on that exact ground ^[1]. On the OCII Western Addition Area 2 footprint the portfolio records 21 subsidized developments holding 914 units, 821 of them income-restricted, which is 89.8 percent of the total ^[1]. The earliest dated construction on those blocks is 1986, and the bulk of the units arrive in the 2000s and 2010s, two generations after the bulldozers ^[1]. Tenure on the footprint runs near even between rental and ownership while the citywide subsidized portfolio runs about two-thirds rental, and the rebuilt stock skews toward formerly-homeless residents and small units ^[1]. We ran no experiment, interviewed no one, scraped nothing, and built no model. We count no one who was displaced, because a portfolio of buildings holds no record of the people removed before those buildings existed. Every displacement magnitude here is external, borrowed from the literature and from city records and set beside our counts only so the difference in scale can be seen. Against the city's estimate of about 5,893 displaced households, 914 subsidized units now stand, which is roughly one rebuilt unit for every six households the city says were cleared ^[1][10]. The rebuild reads as slow and partial against a removal that was fast and close to total.

What Stands There Now

Walk west from the edge of the Fillmore on a weekday and the ground gives nothing away. Low apartment buildings carry names over their doors. A few townhouse rows sit behind small front gates. There are senior buildings, and blocks where a corner store anchors one end and a community room anchors the other. Some of the brick and stucco is plainly newer than the rest, and the decades sit next to one another without ceremony. A person who did not know the history would read these blocks as ordinary, the kind of subsidized and mixed housing that fills in over time in an American city.

What that person cannot see is that the land underneath was emptied on purpose. These are the blocks the San Francisco Redevelopment Agency designated for renewal, and much of what stands here now stands on ground cleared two generations ago. The buildings are real. The people in them are real. The public money that built them is real and traceable in a dataset the city publishes. So is the absence underneath, though the buildings cannot show it.

That absence is the reason an honest account has to declare its limits before it counts anything. A reader should never have to wonder where a figure came from, so here is the order of things. This is a synthesis of published scholarship placed beside a descriptive analysis of one public dataset, the San Francisco MOHCD and OCII Affordable Housing Portfolio distributed through DataSF ^[1]. The original contribution is the description. It is modest by design, and it is bounded hard.

Two streams of evidence run through these pages, and they never join. The first is the historical and peer-reviewed record of what urban renewal did, to San Francisco and to the country, and it supplies every displacement magnitude. When you read a number for people or households removed, it comes from a cited author or a city document, marked as such. The second stream is our count of what public subsidy rebuilt on the footprint, and it supplies the 21 developments, the 914 units, the 821 that are income-restricted, and the dates and tenures and unit sizes that go with them ^[1]. We keep the two apart sentence by sentence, because the easiest way to make this paper dishonest would be to let them blur.

One limit governs everything that follows. The dataset is a portfolio of what was built, not a census of who was removed. It records developments, units, tenure, construction dates, and the populations a building is configured to serve. It holds no record of a single household displaced in the 1950s or 1960s, because people cleared from a neighborhood do not appear in a registry of buildings put up decades later. The roughly 5,893-household figure that fixes the scale of the removal is external, borrowed from the city's own accounting by way of reporting and set beside our counts only to show the difference in size between what left and what came back ^[10]. We do not compute it. We cannot check it against our rows. We will not let it pass as something the analysis produced.

The long arc is already visible in two dates. Clearance ran through the 1950s and 1960s. The rebuild we can document on this exact ground is dated mostly to the 2000s and 2010s, the earliest construction in the portfolio reaching back only to 1986 ^[1]. Between the bulldozers and the buildings now standing on the cleared blocks lies a gap of about two generations. That gap is the subject. We will describe it carefully and let the description do what an honest account allows. We will not explain why it is so wide, because the data cannot establish cause, and pretending otherwise would forfeit the one thing the paper has, which is that every figure in it traces to a source a reader can check.

A note on where these habits come from. Rooted Forward is a Chicago civic project, and the discipline it brings to a South Side block carries over to a San Francisco one without much adjustment. Look at the ground first. Name the public actions that shaped it. Keep the people at the center even when the records refuse to count them. Say less than you wish you could, because the alternative is to say more than you know. Each major turn of what follows opens on the footprint or on the human arc, and only then lets the counts speak. The buildings on the cleared blocks are where this started. They are where it will end.

How the Ground Was Cleared

The Western Addition was one product of a federal machine, and the machine comes into focus before the neighborhood does. Its legal engine was the Housing Act of 1949. Title I of that act let cities assemble land, clear it, and resell it to private developers with federal money covering most of the loss between what the land cost to acquire and clear and what it fetched on resale. The statute's stated aim was a decent home and a suitable living environment for every American family. In practice the program fell hardest on the neighborhoods least able to fight back. It ran for a quarter century, from 1949 to 1974, and the historian Brent Cebul reads that quarter century as a systematic teardown of Black neighborhoods, carried out city by city with federal subsidy and local enthusiasm ^[8].

The resale step is the part of the mechanism that often gets lost, and it matters for understanding who the program served. Title I did not simply demolish. It cleared land and then handed it, cheaply, to private developers to rebuild as those developers saw fit, with the federal subsidy covering the gap between the high cost of assembling and clearing occupied urban land and the lower price it fetched once empty ^[8]. The public paid to make the land available. Private interests decided what rose on it, and what rose was frequently not housing for the people who had been removed but commercial space, institutional expansion, or housing priced well beyond their reach ^[8]. That structure is why renewal so reliably reduced the stock of housing available to the displaced even as it was sold as housing policy. The subsidy flowed to clearance and to the developers who took the cleared land. The households cleared off it were left to find somewhere else, usually in the next district the color line still permitted, which concentrated them further and set up the following round.

The arithmetic Cebul reports defines the era. Black Americans were about 13 percent of the national population in those years. They were at least 55 percent of the people displaced by renewal ^[8]. The program had a folk name at the time, Negro removal, and Cebul takes that name seriously as a description of what the policy accomplished rather than as a slur shouted from the sidelines ^[8]. The phrase was not invented by critics looking back. It was current while the bulldozers ran, a plain-language summary by the people watching their blocks come down.

The national scale is hard to hold in the mind, which is part of why the work of assembling it matters. Mindy Thompson Fullilove, a psychiatrist and public-health scholar, gathered the federal record into figures a reader can carry. On her accounting, urban renewal affected roughly 1,600 African American neighborhoods and displaced on the order of one million people, through some 2,500 distinct projects spread across 993 cities ^[3]. About 75 percent of the people displaced were people of color ^[3]. Those numbers describe a national reordering of where Black and brown Americans were allowed to live, executed under the banner of improvement and paid for from Washington.

Fullilove's contribution runs past the count. She insists the count understates the thing. What renewal destroyed was the dense web of relationships and routines that a neighborhood is, and a tally of demolished structures cannot register that. She gave the loss a clinical name, which we come to when we reach the Western Addition itself. For now the point is the arithmetic, and the arithmetic is hers, not ours.

A second assembly of the federal record sharpens the picture by working straight from the project files. The Renewing Inequality project, led by Robert K. Nelson and colleagues at the University of Richmond, digitized federal Urban Renewal Project records and mapped family displacements city by city for the years 1955 to 1966 ^[7]. That dataset documents more than 300,000 families displaced across more than 600 cities in that eleven-year window alone, with the burden falling disproportionately on families of color ^[7]. The value of the project lies in its granularity. It pins displacement to particular cities and particular projects rather than leaving it as a national aggregate, so a reader can see that renewal was a routine of hundreds of clearances running at once in cities large and small, not a handful of dramatic ones. The window is narrower than Fullilove's full-program estimate, which is one reason the two figures sit at different magnitudes. They are not in conflict. They describe the same machine from two vantage points, one taking in the program's whole arc and one drawn from the granular paperwork of its most active decade. Neither figure is ours. Both stand behind a single plain claim, that postwar renewal moved hundreds of thousands of mostly nonwhite families off their land with public money.

The two counts vary for a reason, and a reader who notices the spread between roughly one million people in Fullilove's full-program figure and more than 300,000 families in the Renewing Inequality window might reasonably ask which is right. Both are. They count different things over different spans. Fullilove's estimate runs across the whole 1949-to-1974 life of the program and counts people ^[3]. The Richmond figure runs from 1955 to 1966 and counts families, a unit that bundles several people into one record ^[7]. A figure in people and a figure in families are not the same measurement, and neither is a figure for a quarter century the same as a figure for eleven years. We lay the two side by side rather than reconcile them into a single number, because reconciling them would require assumptions about household size and program intensity that belong to the original researchers, not to us. The honest summary is the one both sources share. The program displaced a great many mostly nonwhite households, and it did so as a matter of routine federal practice rather than as a scatter of exceptions.

A question hangs over all of it. Why did some cities clear so much more than others, and why did the clearing concentrate where Black families had recently arrived? The answer that the research supports is causal, and our own data supports nothing of the kind, so the claim has to belong entirely to the people who established it. Ying Shi, Daniel Hartley, Bhashkar Mazumder, and Aastha Rajan studied exactly this, using a shift-share instrumental-variables design to isolate the effect of Black in-migration during the Great Migration on the renewal decisions local governments made ^[6]. Their finding, theirs and not ours, is that larger Black in-migration led local governments to undertake more urban renewal projects and to displace more families ^[6]. The Migration brought Black families north and west. The receiving cities, on this evidence, answered by clearing more ground and removing more households.

The design is what separates their causal claim from the descriptive work here, and it bears a brief description. A simple correlation between Black in-migration and renewal activity would prove little, since cities that drew migrants differed from cities that did not in dozens of ways that might independently drive clearance. A shift-share instrument addresses that by predicting each city's Black in-migration from forces outside the city's own later choices, roughly the size of a city's earlier Black community combined with national migration flows out of the South, and then using only that predicted, externally driven variation to estimate the effect on renewal ^[6]. The method is built to defend a sentence with the word led in it. Our portfolio supports no such sentence. We can say what was built on the footprint and when. We cannot say that any condition caused any outcome, and the contrast with Shi and colleagues is the contrast between a study designed for causation and a description that declines it. We report their result in their names because it is theirs, and because the premise of our own work is that a portfolio of buildings cannot manufacture cause.

San Francisco was not a special case running against a national grain. It was one instance of a national program that did the same kind of thing in hundreds of places, with the same disproportionate weight on Black and other nonwhite neighborhoods. The 1949 Act handed cities the tool. Federal money made the tool affordable. Local coalitions chose where to swing it. And the people who lived on the most valuable, least defended ground, very often Black people who had arrived during the Migration, are the people the records show being moved. The literature carries the magnitudes and the causation here, because the magnitudes and the causation are the literature's to carry. Our own work waits until the ground is cleared and the rebuild begins, which is the order in which it actually happened.

A Pattern That Ran Through Many Cities

Rooted Forward does most of its work in Chicago, and the choice to read San Francisco here is a comparative one, made because the same federal machine ran through both. Chicago is one of the cities inside the national records the historians and the Richmond project assembled. It sat among the 993 cities Fullilove counts and among the more than 600 cities in the Renewing Inequality window, and the Great Migration that Shi and colleagues tie to renewal intensity brought hundreds of thousands of Black Southerners to Chicago in the same decades it was reshaping the receiving cities of the West Coast ^[3][7][6]. The neighborhoods that renewal cleared in Chicago and the neighborhoods it cleared in San Francisco were produced by the same combination of forces, a Black population concentrated by the color line into a few districts, land that became valuable as cities grew around those districts, and a federal program that paid local coalitions to clear and rebuild. The particulars differ. The structure does not.

The comparison earns its keep because it guards against a misreading that any single-city study invites, that what happened in the Western Addition was a local accident of San Francisco's politics. Chester Hartman's account makes San Francisco's downtown coalition vivid and specific, and a reader could close his book thinking the clearance turned on the personalities and pressures of one city's establishment ^[2]. The national record corrects that. The same outcome appeared in city after city run by different coalitions answering to different local pressures, which is the signature of a national policy rather than a local quirk ^[8][7]. San Francisco's politics shaped how the clearance happened on these particular blocks. The fact that clearance happened at all, and fell where it fell, is a national fact that Chicago and dozens of other cities share.

The serial-forced-displacement framework gives the comparison its sharpest edge, and it is the framework that most directly connects a Chicago project to a San Francisco footprint. Fullilove and Wallace describe the same neighborhoods enduring wave after wave, redlining and then renewal and then deindustrialization and then disinvestment, each one compounding the last across most of a century ^[5]. That sequence is not a San Francisco story or a Chicago story. It is the shape of what happened to formerly redlined, majority-Black districts in American cities generally, and it is why a civic project rooted in one city can read another city's cleared ground and recognize the pattern in it ^[5]. The Western Addition was redlined before it was renewed, and the same is true of the South and West Side neighborhoods that Rooted Forward's other work documents. We do not import any Chicago number into the San Francisco analysis, and we attach no figure to this comparison that the cited national sources do not carry. The comparison is structural, a way of seeing the footprint as one instance of a pattern, and the numbers that follow are San Francisco's alone.

There is a reason to read one footprint closely rather than only at the national scale, and it is the same reason the field exemplars in this kind of work return to single blocks and single families. A national figure, a million people displaced or 300,000 families, establishes the magnitude but loses the texture, the way a particular neighborhood was assembled, cleared, and slowly rebuilt over decades on a piece of ground a reader could stand on ^[3][7]. The Western Addition affords that close read because the record on it is unusually full, anchored by Hartman's political history and Brahinsky's project-level reporting and now by a public portfolio that lets the rebuild be counted ^[2][9][1]. Reading it closely is a way of testing the national pattern against one well-documented instance, of asking whether the abstraction holds up when you bring it down to a single set of blocks. It does hold up here, which is part of what makes the case worth telling. The clearance fell where the pattern predicts, on a Black and immigrant district concentrated by the color line, and the rebuild came late and partial in the way the displacement literature would lead one to expect. The single case does not prove the national pattern, since one instance never proves a generalization. It illustrates the pattern with a concreteness the aggregates cannot reach, and it does so on ground a reader can still walk.

The Western Addition in Particular

Before renewal the Western Addition was one of the densest and most mixed neighborhoods in San Francisco. It had absorbed Japanese American residents before the Second World War. During the war, when internment forced those residents out of their Victorian flats, Migration-era Black newcomers filled the same buildings, drawn to one of the few districts in the city where the color line and restrictive covenants did not bar them. By the late 1940s the neighborhood held a Black cultural and commercial life dense enough to earn the Fillmore its later reputation as a center of West Coast jazz. It was the kind of place postwar planners looked at and called blight, and the kind of place the people who lived there called home. What happened to it is documented in unusual detail, and that documentation is what lets us treat this footprint as a real place with a real history rather than as an abstraction on which a national pattern gets projected.

The neighborhood's layered history makes it a distinctive case even within the national pattern, and the layering is exactly what the serial-displacement framework was built to see. One population was removed by force in the 1940s, the Japanese Americans sent to internment camps, and another population, Black Southerners arriving in the wartime Migration, moved into the housing the first had been compelled to vacate, hemmed into the Western Addition because the color line closed most of the rest of the city to them ^[5]. Then, beginning in the 1950s, the second population was cleared in turn by renewal ^[9]. The same blocks absorbed two distinct upheavals to two different communities inside a single generation, which is the serial pattern Fullilove and Wallace describe rendered unusually visible on one piece of ground ^[5]. We do not collapse the two episodes into one, because internment and urban renewal were different policies with different victims and different mechanisms. We name the sequence because it shows the ground itself carrying a history of repeated removal, and because that history is the backdrop against which the slow, late, partial rebuild has to be read. The blocks were not a blank slate that renewal happened to clear. They were a site where displacement had already happened once before it happened again.

The word blight did the legal work that made the clearance possible. Title I let cities clear land they could designate as blighted, and the designation was the hinge on which condemnation turned. A neighborhood declared blighted could be assembled and demolished. The trouble, which the redevelopment literature returns to again and again, is that the same conditions that earned a Black neighborhood the label, age, density, overcrowding, deferred maintenance, were conditions the color line itself had produced, by packing a growing Black population into the few districts open to it and starving those districts of the mortgage credit and public investment that flowed freely elsewhere ^[2][8]. The density planners read as blight in the Western Addition was the density that segregation had forced. The overcrowding was a symptom of the very exclusion that then justified the clearance. So the designation operated as a kind of circular machinery. Exclusion produced the conditions, the conditions earned the label, and the label authorized the removal of the people exclusion had concentrated there. Hartman's account of San Francisco and Cebul's account of the national program both turn on this circularity, the way a status manufactured by discrimination became the legal pretext for displacement ^[2][8]. We name it here because it explains how a neighborhood full of homes, businesses, and a settled community could be cleared as a slum by an administrative measure that counted none of those things.

The anchor account is Chester Hartman's. In City for Sale, Hartman traces the politics of San Francisco redevelopment across decades, and his central finding is that clearance was chosen, not suffered as an accident of policy. A downtown business establishment worked through and with city government to remake the central neighborhoods in an image that served downtown's interests, and the Western Addition sat close to the commercial core on land that coalition wanted reordered ^[2]. Hartman's book is the definitive telling of who decided, who profited, and who was moved ^[2]. The deciding happened in rooms, by people with names, against residents who organized, litigated, and largely lost. That is the political shape of what the bulldozers then carried out, and Hartman is the reason we can describe it as a decision rather than a misfortune.

The strength of Hartman's account, for the purposes of this paper, is that it refuses the language of inevitability that often settles over urban history. Renewal is easy to narrate as a thing that happened to cities, as though clearance were weather. Hartman insists on agents. He tracks the alliance between a growth-minded business leadership and a redevelopment agency armed with federal money and the power of eminent domain, and he shows that the same machinery which cleared the Western Addition was the machinery behind the larger remaking of the city's core ^[2]. The residents of the neighborhood were not passive in this. They organized, they brought litigation, and in some instances they slowed the agency down. They did not stop it. Hartman's record of that asymmetry, of organized residents losing to better-funded institutions holding the legal whip of condemnation, is what lets a reader understand the clearance as the outcome of a contest rather than the absence of one ^[2]. The blocks our analysis sits on came down because a coalition with the power to take land decided they should, over the objection of the people who lived there.

The granular record of the clearance itself comes through Rachel Brahinsky's reporting. Renewal of the Western Addition came in two designated projects, A-1 and A-2, and Brahinsky documents both. A-1, the earlier and smaller, displaced roughly 4,000 people by the mid-1960s ^[9]. A-2, far larger, displaced somewhere between 10,000 and 13,000 people by 1970 ^[9]. Those figures are hers, drawn from her long-form account of how redevelopment tore through the neighborhood, and we attribute them to her precisely because they are the kind of number that matters most and that our data cannot touch ^[9]. A-2 is the project whose footprint our analysis sits on. When the later sections count what public subsidy rebuilt on the OCII Western Addition Area 2 footprint, they count it on the very ground Brahinsky records as having displaced ten to thirteen thousand people inside a single decade. Her figure and our count stay in separate hands. The geography is the same geography, and that sameness is the whole reason the comparison can be made at all.

The two-project structure matters for understanding the timeline our own data later picks up. A-1 and A-2 were not a single clean sweep but a sequence, the first project clearing one set of blocks and the second, larger and later, extending the clearance across a much wider area into 1970 ^[9]. Brahinsky's account runs the A-2 displacement out to 1970, well over a decade after the project began, which already marks the clearance itself as a prolonged process rather than a single event ^[9]. That extended timeline is part of the historical record, not something we discovered, and it bears on how to read the gap our portfolio shows. By the time the earliest subsidized construction we can date arrives in 1986, the A-2 clearance Brahinsky documents was a decade and a half or more in the past ^[9][1]. The displacement was not a single moment but a drawn-out process running into 1970, and the subsidized rebuild we can date is a separate and even more drawn-out process beginning in the mid-1980s. Holding both timelines in view, Brahinsky's for the clearance and ours for the subsidized rebuild, is how the forty-year arc comes into focus, each timeline sourced to where it actually comes from.

The city has since produced its own reckoning, and Eleni Balakrishnan's reporting brings it into view. The figure the city now uses for households displaced by the San Francisco Redevelopment Agency in the Western Addition is roughly 5,893, most of them Black ^[10]. That is a household count, where Brahinsky's are people counts, which is part of why the two sit at different magnitudes and why we never collapse one into the other. It is also the city's own number, an institution tallying the cost of what an earlier version of itself did ^[10]. Out of that reckoning came a policy instrument meant to address the harm, the Certificate of Preference, designed to give displaced households and their descendants a priority claim on the city's affordable housing so that some path back to the neighborhood might exist ^[10]. Hold the program in mind. How little it has been used returns when our own counts are on the table, because the distance between a promise of return and the practice of return is part of what the footprint now shows.

Numbers like 5,893 households or 13,000 people can flatten into statistics, and the literature insists they should not. Fullilove gave the human cost a name in the peer-reviewed paper where the idea first appeared. In her formulation, root shock is the traumatic stress reaction a person suffers on losing what she calls the emotional ecosystem in which a life is organized, the web of relationships, routines, landmarks, and belonging that a neighborhood holds, when that neighborhood is demolished around them ^[4]. The concept recasts displacement as an injury rather than an inconvenience, a wound to the individual and to the collective body of the community at the same time ^[4]. Fullilove drew the term from horticulture, where root shock names what happens to a plant wrenched out of its soil, and the borrowing is exact. A neighborhood is the soil, and the people pulled out of it are not simply relocated. They are damaged by the pulling ^[4]. When the Western Addition lost its blocks it did not simply lose addresses. The people lost the web of relationships and familiar ground their lives ran on, and root shock is Fullilove's term for what that loss does to a person ^[4]. We bring the concept in here, with its source, because the rest of the paper deals in units and decades, and units and decades stay legible only if the human stakes underneath them stay in view.

There is a further layer the literature adds, and it unsettles any tidy story of harm followed by repair. Fullilove and Rodrick Wallace document what they call serial forced displacement, the pattern by which the same neighborhoods endure not one upheaval but a sequence of them across the better part of a century, redlining and then renewal and then deindustrialization and then disinvestment, each wave compounding the last ^[5]. The Western Addition fits the pattern. It was redlined before it was renewed, and it has been pulled by disinvestment and reinvestment pressures since. So the rebuild the later sections describe does not arrive on neutral ground that was cleared once and left to wait. It arrives on ground worked over repeatedly, by public and private hands, in ways the serial-displacement framework helps a reader keep in view ^[5]. That framework is theirs. The magnitudes of the original removal are Brahinsky's and the city's. What is ours starts now, on this footprint, with the buildings public subsidy put back.

Reading the Portfolio, and What It Cannot Say

The source is a single dataset, the San Francisco MOHCD and OCII Affordable Housing Portfolio, published by the City and County of San Francisco through its Mayor's Office of Housing and Community Development and its Office of Community Investment and Infrastructure, and distributed through DataSF as Socrata dataset aaxw-2cb8 ^[1]. It ships in the repository behind this paper as two CSV files, and the analysis is reproducible on the site, so any reader can run the same rows we ran ^[1]. We did not collect the data. The city did. We filtered it and described it, and the limits of that description matter enough to state before any result, because the limits govern the results.

A word on what the dataset actually is, since the name can mislead. It is a portfolio, a record of subsidized developments the city tracks, with one row per development carrying its name, address, neighborhood, OCII project area, sponsor agency, tenure, project status, development type, an estimated or actual construction date, a total unit count, a count of income-restricted units, and rollups of senior, formerly-homeless, and bedroom-size units. The income-restricted count is the number of units in a development that carry a legal cap on the income of the household that may rent or buy them, the mechanism that makes a unit affordable in policy terms rather than merely cheap. On the footprint, 821 of the 914 units carry that restriction, which is 89.8 percent, so the rebuilt stock is overwhelmingly income-restricted rather than market-rate units that happen to sit in subsidized buildings ^[1]. What the dataset is not is a history. It does not reach back before the developments it lists, it carries no record of what stood on each site before, and it holds nothing about the households that any development replaced. Every field describes the building that exists now. None describes the ground's past. That single property is the source of the central limit we keep returning to, and it is built into the structure of the data rather than a gap we could fill with more careful reading.

The most consequential choice was how to define the footprint, and it belongs in the open rather than buried in a method note. The portfolio carries a field naming the OCII project area each development sits in. We treat the renewal footprint as the developments where that field reads Western Addition Area 2, and only those, which yields 21 developments ^[1]. A larger geography was available. The Western Addition planning neighborhood is broader than the OCII renewal area and contains 94 developments in this portfolio ^[1]. We report that broader number separately and never let it stand in for the footprint, because the planning neighborhood is not the renewal area. The blocks the Redevelopment Agency actually cleared, and that OCII still tracks as Western Addition Area 2, are the blocks our headline counts describe. Widening the lens to the whole neighborhood would inflate the rebuild and blur the correspondence between cleared ground and rebuilt ground that the paper rests on. The footprint is the OCII area, 21 developments. The neighborhood figure stays at arm's length.

The analysis behind the counts is deliberately simple, and its simplicity is a feature. Every figure in these pages comes from filtering and tallying the rows of two CSV files, with no statistical modeling and no estimation. The footprint counts come from selecting the rows where the OCII project-area field reads Western Addition Area 2 and summing their unit columns. The citywide comparators come from the full 849-row portfolio. The decade timeline comes from parsing the construction-date field and bucketing by ten-year period. The tenure, development-type, and bedroom breakdowns come from grouping on the relevant fields and counting. Anyone with the two files and a spreadsheet can reproduce the headline numbers, and the files ship alongside this analysis so they can do exactly that ^[1]. We checked the arithmetic rather than trusting it, confirming that the bedroom rollups account for 792 of the footprint's 914 units, that all 21 developments carry a parseable construction date, that the income-restricted count of 821 and the total of 914 have no missing values, and that the 325-row developments file is fully contained in the 849-row portfolio ^[1]. The point of saying all this is not to dress a simple count in the costume of rigor. It is the opposite. The work is a careful read of a public file, the kind any reader could redo, and a paper that claims no more than that owes its readers the means to check it.

For scale, those 21 developments sit inside a larger OCII universe. The developments file records 86 developments across all nine OCII project areas, holding 5,448 units in total, of which 4,967 are income-restricted ^[1]. Ranked by units, the areas run Mission Bay first at 1,890 units across 16 developments, then Western Addition Area 2 at 914 across 21, then Bayview Hunters Point at 766 across 9, South of Market at 464 across 7, Rincon Point-South Beach at 445 across 5, Candlestick Point at 339 across 5, Yerba Buena Center at 327 across 4, the Hunters Point Shipyard Phase 1 at 183 across 18, and Transbay at 120 in a single development ^[1]. Western Addition Area 2 is the second largest of the nine by unit count and the largest by number of developments. So the footprint is not a marginal slice of the OCII portfolio. It is one of its heaviest concentrations of subsidized housing, which is part of why it rewards reading on its own. The ranking carries a limit, though. A high unit count on the Western Addition footprint is not evidence of a strong rebuild relative to need, because the need on that footprint, the scale of the original removal, is not something the OCII totals measure. Mission Bay and Mission Creek were largely rail yards and industrial land before redevelopment, not cleared residential neighborhoods, so a unit count there sits against a very different history than the same count on ground where thousands of households once lived. The areas are comparable as portfolios. They are not comparable as repairs, and we do not treat them as such.

A structural fact about the two files governs every comparison, and we verified it directly rather than assuming it. The developments file holds 325 rows. The citywide portfolio file holds 849 rows. Every one of the 325 development IDs in the first file appears in the 849-row portfolio, with none missing, which means the developments file is a strict subset of the citywide file and not an independent sample drawn alongside it ^[1]. We confirmed the containment by checking the IDs, and the consequence is methodological. We use the full 849-row portfolio as the citywide comparison universe, and we treat the two files as what they are, a whole and a part of that whole ^[1]. When a later section says the footprint compares thus and so to the citywide portfolio, the citywide portfolio includes the footprint inside it. That is the honest framing, and it slightly compresses any contrast we report, since the part is being measured against a whole that contains it. The compression cuts against our own headline differences, not toward them, so where a contrast survives it survives despite the overlap.

Several caveats bound the numbers that follow, and each one deserves a plain sentence before a reader leans on a figure. The files are point-in-time snapshots of a live dataset the city updates, so the totals will drift as developments are added, completed, or reclassified ^[1]. Anyone who pulls aaxw-2cb8 a year from now should expect slightly different counts, which is a property of the source and not an error in the reading. The senior-units field is sparsely populated, recording any senior units for only 9 of 849 developments citywide and only 1 of 21 on the footprint, so it badly understates senior housing ^[1]. We exclude it from the charts and report it only where transparency demands. The bedroom-size rollups do not sum to total units, because treatment beds and unspecified units are never bucketed into bedroom categories, so on the footprint the bedroom counts cover 792 of 914 units, which is 86.7 percent of the total ^[1]. Every bedroom share we report is therefore a share of total project units and does not add to 100 percent. We say so each time rather than quietly normalize the gap away.

One caveat carries extra weight for the question of timing, so it goes here, before the reader meets the decade counts. The construction-date field records estimated or actual construction, and it includes future and in-progress dates, the latest reaching to February 2027 ^[1]. The practical effect is that the most recent decade bucket counts developments that are not yet finished, so the 2020s figure includes work that is dated but still underway ^[1]. These are construction dates, not certificate-of-occupancy dates, which means the timeline marks when building is recorded as happening, not when residents could move in ^[1]. We surface this rather than let a late-decade number read as more settled than it is.

The choice to describe rather than to infer is a methodological position, and we defend it rather than apologize for it. A portfolio of buildings can answer a narrow set of questions cleanly. How many developments sit on the footprint, how many units they hold, what share is income-restricted, when each was built, how they are configured. Those are questions of counting, and the data answers them with no inference required beyond reading the rows correctly. The portfolio cannot answer the questions a reader most wants answered. Whether the rebuild repaired the harm, whether the displaced or their descendants came back, why the rebuild took as long as it did. Those are questions of cause and of biography, and the data contains neither causes nor biographies. A weaker paper would reach for them anyway, dressing a count in the language of explanation. We hold the line at description because description is what this source can support honestly, and a true small claim is worth more than an impressive claim the evidence cannot carry. The counts that follow are offered in that spirit, as accurate answers to the questions the data can actually answer.

One restatement of the central limit belongs here, where it can discipline the comparison that follows. This portfolio counts what was built. It does not count who was removed. Nowhere in the 21 developments, or in the 849 citywide rows, is there a record of a household displaced in the 1950s or 1960s, because the displaced do not appear in a registry of buildings constructed decades after their removal ^[1]. Every displacement figure in this paper is external, drawn from the literature and from city records, placed beside our counts only to frame scale. With that boundary fixed, the counts can speak. The first thing they say is when the rebuild arrived.

A Rebuild That Arrived Late

Set two dates side by side and the long arc announces itself. The Western Addition was cleared in the 1950s and 1960s. The earliest subsidized construction dated anywhere on the OCII Western Addition Area 2 footprint, in the city's own portfolio, is 1986 ^[1]. Between the emptying of the blocks and the first dated public-subsidy building the dataset records on them lies the better part of three decades, and the weight of the rebuild lies later still, in the 2000s and 2010s ^[1]. That is the finding at the center of the paper's long-arc claim, and it is a finding about timing and nothing else. We can say when the recorded rebuild happened. We cannot say from this data why it took as long as it did, and we will not try.

The by-decade counts make the shape concrete. All 21 developments carry a parseable construction date, with none blank or unreadable, so the timeline rests on the complete set rather than a fraction of it ^[1]. The 1980s contribute 100 units across three developments. The 1990s add 145, also across three. The 2000s bring 121 across five developments. The 2010s account for 405 units across eight developments, by a wide margin the heaviest decade. The 2020s show 143 units across two developments ^[1]. The 2010s alone hold more units than the 1980s, 1990s, and 2000s combined, 405 against a combined 366 ^[1].

Read the curve plainly. Nothing on this footprint predates the mid-1980s in the dataset's record, which already places the start of the documented rebuild a full generation after the clearance. The rebuild then proceeds unevenly, a modest 1980s and 1990s, a smaller 2000s, and a pronounced concentration in the 2010s. The decade that dominates the rebuild begins half a century after the neighborhood was cleared. Whatever else the curve shows, it shows that the public-subsidy answer to the Western Addition's clearance, as far as this portfolio records it, is overwhelmingly a twenty-first-century answer to a mid-twentieth-century action.

The unevenness inside the curve calls for the same restraint. The line does not climb steadily. It rises a little in the 1980s, rises again in the 1990s to 145 units, then falls back in the 2000s to 121, then jumps to 405 in the 2010s, then drops to 143 in the partial 2020s ^[1]. A reader looking for a tidy story of accelerating repair will not find one in that shape. The 2000s are lower than the 1990s, which is the opposite of acceleration, and the 2010s spike sits alone rather than as the top of a rising slope. We will not build a narrative on the dips and the spike, because a single development landing in one decade rather than the next can move a small annual count sharply, and with only 21 developments spread across five decades the per-decade figures are coarse. The honest reading is the coarse one. Construction on this footprint is a recent and lumpy phenomenon, concentrated in the 2010s, essentially absent before the mid-1980s, and the lumpiness warns against reading too fine a trend into it. The shape is real. The smoothness a reader might want to impose on it is not in the data.

The gap between removal and rebuild therefore spans roughly two generations. Clearance fell in the 1950s and 1960s. The bulk of the rebuild falls in the 2000s and 2010s. We describe the span as a pattern in the numbers and resist the pull to explain it. Candidate explanations are easy to list and impossible to adjudicate from this data. Funding cycles turned over. Litigation dragged. Nonprofit development capacity took years to assemble. Federal housing policy shifted under the program more than once. Building anything in San Francisco is slow and expensive on its own terms. The portfolio settles none of these against the others. To name a cause would be to claim something the data cannot support, so the timeline says what it can, that the rebuild arrived late and arrived mostly recently, and the why is left to work that has the evidence for it.

The asymmetry between the two timelines is itself a finding, and it sharpens what the long arc means. Clearance was compressed. Brahinsky records the A-2 displacement running from the late 1950s into 1970, a removal of ten to thirteen thousand people inside little more than a decade, which is fast for an undertaking of that size ^[9]. The rebuild is diffuse. Our portfolio spreads 914 units across five decades, with no single decade before the 2010s topping 145 units and the heaviest decade still holding under half the total ^[1]. A removal that took a decade and a half has been answered, so far, by a rebuild stretched across four decades and counting. Demolition is quick because it requires only the power to take and clear land, which the Redevelopment Agency held in concentrated form ^[2]. Rebuilding is slow because it requires money, sponsors, approvals, and the assembly of each development one at a time, none of which arrives on the schedule that clearance did. The portfolio cannot prove that this is why the rebuild diffused, and we do not claim it does. But the contrast between a compressed removal and a diffuse rebuild is visible in the two records laid together, Brahinsky's and ours, and it is the temporal core of what forty years on cleared ground actually describes ^[9][1].

The 2020s figure of 143 units needs its caveat fastened to it, the same one flagged before the chart, because a reader could otherwise take it as completed work standing on the ground today. It is not all standing yet. The construction-date field includes future and in-progress dates reaching to early 2027, so the 2020s bucket counts developments that are dated but not necessarily finished, and these are construction dates rather than certificate-of-occupancy dates ^[1]. The late-decade number is best read as the rebuild recorded as underway, some of it incomplete, rather than as units a household could move into now. The qualification does not soften the larger shape. Even granting the 2020s in full, the arc still runs from a cleared mid-century neighborhood to a rebuild concentrated forty and fifty years on. The forty years of public subsidy in the title is not a figure of speech here. It is a count of decades between the clearance and the buildings, and the count says the answer came late.

Who the New Blocks Were Built For

A timeline says when the blocks were rebuilt. It says nothing about what kind of housing went up or who it was meant to hold. For that, we place the footprint's stock beside the citywide subsidized portfolio along two dimensions, tenure first and then the populations and unit sizes the stock is configured for. Both are descriptive comparisons of distributions, the footprint as a part against the citywide whole that contains it, and neither carries causal weight. We describe the character of what was built. We do not explain how it came to have that character.

Tenure is the simplest cut and the contrast is clean. On the footprint, the 21 developments split almost evenly, 11 ownership and 10 rental, which works out to 52.4 percent ownership against 47.6 percent rental ^[1]. The citywide portfolio leans the other way and leans hard, 569 rental developments against 280 ownership, or 67.0 percent rental to 33.0 percent ownership ^[1]. The city's subsidized portfolio as a whole is about two-thirds rental. The renewal footprint runs close to half and half, with a slight tilt toward ownership.

That near-even split is a real departure from the citywide pattern, and it rewards a close look without over-reading. Subsidized ownership housing is the rarer thing in most affordable portfolios, partly because the public subsidy in ownership models is harder to keep attached to a unit once it sells, and on this footprint it reaches parity with rental, where citywide it sits at a third. We note the contrast and stop. The portfolio does not tell us why the footprint accumulated a heavier share of ownership units, whether through deliberate program choices, the particular developments that happened to land here, or the long sequence of decisions that filled these blocks in across forty years. The distribution is the finding. The reasons behind it sit outside what this data can show.

A second cut, development type, says something about what the rebuild was physically doing on the land. The portfolio sorts each development into new construction, rehabilitation of an existing building, or adaptive reuse of a structure built for another purpose. On the footprint, 14 of the 21 developments are new construction and 7 are rehabilitation, which is 66.7 percent new against 33.3 percent rehab, with no adaptive reuse recorded ^[1]. Citywide the mix is more even and more varied, 479 new construction, 355 rehabilitation, and 15 adaptive reuse, or 56.4 percent new, 41.8 percent rehab, and 1.8 percent adaptive reuse ^[1]. The footprint tilts harder toward new construction than the city does. That tilt is consistent with the simple fact of the ground, which is that renewal had cleared it. A block that has been demolished down to the lot line is a block you build new on, not one you rehabilitate, because the building that might have been rehabilitated is the building the bulldozers took. We offer that reading as a plausible reading of the distribution and nothing firmer, because the portfolio records the development type without recording why each site got the treatment it did. But the contrast is the kind a cleared footprint would be expected to produce, and the data shows it.

Who the stock serves, and how large its units are, is the other face of it, and here the footprint differs from the city in a consistent direction. Formerly-homeless units make up 24.3 percent of the units on the footprint, 222 of 914, against 11.6 percent across the citywide portfolio, more than double the citywide share ^[1]. The smallest units, single-room-occupancy plus studio, are 39.9 percent of footprint units against 36.0 percent citywide, a slimmer gap in the same direction ^[1]. Family-size units, those with two bedrooms or more, run the other way, only 20.9 percent on the footprint against 32.6 percent citywide ^[1]. The footprint holds proportionally more deeply subsidized small units and proportionally fewer family-size units than the city's subsidized stock as a whole.

The full bedroom breakdown on the footprint fills the shape in. Of the 792 footprint units that fall into a bedroom category, 106 are single-room-occupancy units, 259 are studios, 236 are one-bedrooms, 109 are two-bedrooms, 61 are three-bedrooms, 20 are four-bedrooms, and a single unit is a five-bedroom ^[1]. As shares of all 914 units, that is 11.6 percent SRO, 28.3 percent studio, 25.8 percent one-bedroom, 11.9 percent two-bedroom, 6.7 percent three-bedroom, 2.2 percent four-bedroom, and a fraction of a percent at five ^[1]. More than half the footprint's units, counting SRO through one-bedroom, are built for one or two people. The citywide portfolio carries a heavier middle, with two-bedrooms at 21.1 percent against the footprint's 11.9 percent, which is most of where the family-size gap opens ^[1]. A household with two or three children has fewer doors to knock on among the subsidized units on these particular blocks than among the city's subsidized stock at large. That is a statement about the configuration of the rebuilt housing, not about who lives in it, which the portfolio does not record.

There is a quiet tension in those two facts laid together, and it is the kind of thing a descriptive read can point at without resolving. The Western Addition that renewal cleared was, by every account in the literature, a neighborhood of families. It held households in flats and rooming houses, the dense domestic life of a district where the color line had concentrated a growing Black population ^[2][9]. What public subsidy has rebuilt on those same blocks is configured more for single adults and small households than the citywide stock is, with more than half its units built for one or two people and a thinner band of family-size units than the city carries on average ^[1]. We do not claim the rebuild was designed to exclude families, and the data could not support that claim if we made it, since the portfolio records unit sizes without recording intent or the policy histories of individual developments. The deep-need tilt likely reflects real and pressing needs the city was meeting, the homelessness crisis chief among them, which formerly-homeless units at more than double the citywide share plainly speak to ^[1]. Both things sit in the data at once. The ground that once held families was rebuilt, in part, for a different and smaller kind of household, and the reasons belong to a policy history the portfolio does not contain. We let the juxtaposition stand because it is accurate, and we decline to narrate it into a verdict because the evidence for a verdict is not here.

One honest exclusion belongs here. The senior-units field is too sparsely reported to chart, recording any senior units for only 1 of the 21 footprint developments and 9 of 849 citywide, so any senior share computed from it would understate senior housing and mislead more than inform ^[1]. We leave senior housing out of the comparison rather than present a figure we do not trust. The bedroom shares, as noted, are shares of total project units and do not sum to 100 percent, since 792 of the 914 footprint units fall into bedroom buckets and the rest are treatment beds or unspecified ^[1]. Both limits sit on the chart so it is read for what it is.

The distributions together describe a rebuilt subsidized stock configured toward deep need and smaller households relative to the rest of the city's subsidized portfolio. More formerly-homeless units. More single-room and studio units. Fewer family-size units. That is a description of the stock as it stands, the kind of statement this data can carry. A reader may want to think about why subsidized stock comes to look one way rather than another, and there is recent work that helps, brought in as context and not as a claim about these particular blocks. Fiona Kauer, Elena Lutz, and David Kaufmann find, in peer-reviewed research, that residents displaced by for-profit redevelopment tend to relocate into lower-income neighborhoods and into older, smaller housing, while non-profit-led redevelopment yields milder displacement outcomes ^[11]. The point of their work, for a reader of this paper, is that the sponsor of a redevelopment shapes who ends up where. A profit-driven project and a non-profit project can both clear and rebuild the same block and leave the displaced in very different places, which means the institutional character of who does the building is not a detail but part of the outcome ^[11]. That makes the tenure and configuration of subsidized stock worth attending to, since it is one trace of who built it and for whom.

We cite the finding as theirs and offer it only as a lens for thinking about sponsor and tenure mix in subsidized housing generally ^[11]. It is not evidence about the Western Addition footprint, and we do not stretch it into any. Our portfolio does not record, for each development, whether a non-profit or a for-profit sponsor built it, and even if it did, the Kauer study concerns where displaced residents relocate, which our data cannot observe at all. So the connection is conceptual and stays conceptual. The composition we have shown, deep-need and small-unit heavy, is what our data describes. The interpretive literature stands beside that description, in its authors' names, helping a reader hold it in context without converting a distribution into a cause.

Set the Rebuilt Beside the Removed

Here the two streams meet in one frame, and the discipline of the whole paper comes down to how carefully they are allowed to sit together. On the OCII Western Addition Area 2 footprint, public subsidy has put up 914 units, 821 of them income-restricted, on the blocks renewal cleared ^[1]. Against that stands the city's own external estimate of roughly 5,893 households displaced from the Western Addition by the Redevelopment Agency ^[10]. Divide the one by the other and you get about 0.16 subsidized units rebuilt on the footprint for every household the city estimates was removed, which is the same as saying one rebuilt unit for roughly every six displaced households ^[1][10]. That ratio is the comparison everything else has set up, and it is the one most easily abused, so it gets handled with both hands.

The chart places four quantities side by side, two of them ours and two of them not, and the difference in provenance is the most important thing about it. The 914 total units and the 821 income-restricted units are counts from our reading of the DataSF portfolio ^[1]. The roughly 5,893 displaced households and the roughly 250 Certificate-of-Preference holders housed between 2014 and 2024 are external figures, drawn from city records and reporting, set into the same frame only so the magnitudes can be seen against one another ^[10]. The bars are built to read as two kinds of thing. We are not summing them, not dividing a population by a population of the same type, not treating them as commensurable measurements of one process. We are showing a small internal count next to a large external estimate and letting the size difference land.

The external figures are external, and saying so twice is deliberate, not redundant. The 5,893-household displacement estimate and the 250 Certificate-of-Preference holders housed across that decade come from the city's reckoning and from reporting on it, not from anything we computed ^[10]. Our dataset holds no record of the people cleared in the 1950s and 1960s, and no count of who has returned under the Certificate of Preference. It is a portfolio of buildings. The people who were removed, and the descendants who hold a paper preference to come back, live in a record this dataset does not contain. We borrow their numbers from those who do hold the record, mark the borrowing every time, and keep the borrowed figures from passing as our own.

The Certificate of Preference is the city's own attempt to connect the rebuild to the removal, the one bridge between the two streams that this account otherwise keeps apart. Eleni Balakrishnan's reporting describes a program meant to give households displaced by redevelopment, and their descendants, a priority claim on the city's affordable housing ^[10]. On paper it is exactly the instrument a reader might hope for, a way to route some of the rebuilt stock back to the people the clearance removed. In practice, by Balakrishnan's account, it has rarely worked that way. Roughly 250 Certificate holders were housed across San Francisco in the decade from 2014 to 2024, a small number against the thousands of displaced households the certificates were meant to serve ^[10]. The reporting attributes the shortfall to the ordinary frictions that defeat such programs, the difficulty of finding and verifying descendants two generations on, the scarcity of available units, the gap between holding a preference and clearing every other hurdle to an actual unit ^[10]. We report these as Balakrishnan's findings, not ours, because our data sees none of it. The portfolio shows 914 units on the footprint and cannot say whether a single one went to a Certificate holder.

The program matters to the paper's argument even though it sits entirely outside the dataset, because it shows the city itself treating the rebuild and the removal as parts of one story while the mechanism for joining them runs thin. The existence of the Certificate is the city's acknowledgment that the rebuilt housing and the cleared households belong in the same frame. The 250 figure is the measure of how little that acknowledgment has translated into return ^[10]. Both are external. Both come from the reporting we cite. We place them here because the gap they describe, between a promise of return and a quarter of a thousand households actually housed, runs parallel to the gap our own counts describe between what was removed and what was rebuilt. The parallel is not a finding. It is two separately sourced facts pointing the same direction, and we let them point without fusing them.

So the 0.16 figure is a framing ratio and only that. It is the quotient of an internal count and an external estimate. It is not a recovery rate. A recovery rate would claim that for every displaced household, 0.16 of them got a unit back, and the data cannot support that claim by even a little, because we cannot match one rebuilt unit to one displaced household, and many of the 914 units serve populations and needs with no connection to the families removed sixty years ago. Nor is 0.16 a causal accounting, because the rebuild was not constructed as restitution to the displaced, and we have no basis to treat it as if it were. The ratio does one job. It puts the scale of the documented rebuild next to the scale of the estimated removal so a reader can see that the first is a small fraction of the second. Roughly 914 subsidized units now stand where the city estimates about 5,893 households once stood and were cleared.

What this describes is a gap in scale and a gap in speed, not a proven story about why the gap exists, and that distinction has to stay unmistakable. The scale gap is the 914 against the 5,893, a rebuild that is a small fraction of the removal even before anyone asks who the rebuilt units serve ^[1][10]. The speed gap is the arc the timeline already drew, clearance compressed into the 1950s and 1960s against a rebuild stretched across the 2000s and 2010s and still finishing in the 2020s ^[1]. Put the Certificate-of-Preference figure beside both and the partiality sharpens. Roughly 250 holders housed across a full decade is itself a small number against 5,893 displaced households, though that figure too is external and we report it as the city's, not ours ^[10]. The removal was fast and close to total on these blocks. The rebuild, as the data records it, has been slow and partial. That is the comparison at the size the evidence allows, and no larger. The paper sets the rebuilt beside the removed. It does not, and cannot, explain the distance between them.

What the Cleared Ground Tells Us, and What It Withholds

Stand on those blocks again, now that the counts have done their work. The subsidized buildings are still there, the ones the opening asked you to see before any figure carried weight. What has changed is only what we can read into them. They sit on ground the San Francisco Redevelopment Agency cleared two generations ago, and we can now say, from the portfolio rather than from memory, when public subsidy came back to build on it and roughly what it built. The walls are the same walls. The arc they belong to is the thing the paper has tried to draw.

The shape is plain. The literature records a mid-century removal at a scale and speed this dataset cannot touch, the systematic teardown of Black and Japanese American districts under postwar renewal that Cebul, Fullilove, the Renewing Inequality record, and Hartman document, narrowing through Brahinsky's reporting and the city's own reckoning to the Western Addition and its estimated 5,893 mostly Black households ^{[8][3][7][2][9][10]}. Against that, our descriptive read of the DataSF portfolio dates the rebuild on the OCII Western Addition Area 2 footprint mostly to the 2000s and 2010s, decades after clearance, and counts 914 units across 21 developments, 821 of them income-restricted ^[1]. The rebuilt stock leans toward deep need and smaller households, with formerly-homeless units at 24.3 percent of footprint units against 11.6 percent citywide and family-size units at only 20.9 percent against 32.6 percent citywide ^[1]. The framing ratio sits at the end of it, about 0.16 subsidized units per displaced household, a number we have declined to call a recovery rate every time it has appeared and decline to call one here ^[1][10].

Say plainly what the data can carry. It can describe what public subsidy built on the footprint and when, because those are the things the portfolio records, the development, the unit count, the tenure, the construction date. It can hold one count beside another and let a reader see that one is a small fraction of the other. That is the whole of its reach, and it is not nothing. A neighborhood that was cleared and then left, in the public record, to fill back in slowly across forty years is a neighborhood whose rebuild can be measured, and we have measured the part of it that public subsidy left a trace of.

The measurement carries one lesson worth holding onto, even stated this carefully. Public subsidy did return to the Western Addition. The 914 units are real, the 821 income-restricted units are real, and on blocks the city once cleared there now stands a meaningful body of affordable housing serving people in genuine need ^[1]. A reader could close the account there and call it repair. The counts argue against settling for that reading, not by asserting a verdict but by laying the magnitudes side by side. A rebuild of 914 units across forty years, set against a removal the city itself estimates at 5,893 households inside roughly fifteen, is not the inverse of the clearance ^[1][10]. It is a smaller, slower, differently configured body of housing that happens to occupy the same ground. The honest civic lesson is not that nothing was done, since something was, and it is not that the rebuild undid the harm, since the numbers will not bear that weight. It is that subsidized rebuilding on cleared ground, even at the scale the Western Addition received, runs far behind the removal that preceded it in size, in speed, and in whom it serves. That is a finding a reader can take from descriptive counts honestly, and it is the finding this account supports.

Say just as plainly what the data withholds. It cannot identify, count, or follow the people who were displaced, because it holds no record of them, and the 5,893 is the city's estimate, borrowed for scale, while the people behind it appear in no row we read ^[10]. It cannot tell us whether any rebuilt unit went to a displaced household or a descendant, and many of the units plainly serve needs unrelated to the families removed sixty years ago. It cannot establish cause, and we have made no causal claim in our own voice anywhere in these pages. Where cause appears it belongs to the cited authors, as with Shi and colleagues on the Great Migration and renewal ^[6]. The built-not-census limit governs everything and does not loosen at the close.

The return question is not only historical. The Certificate-of-Preference program exists right now, the city's instrument meant to let displaced households and their descendants come back, and by the reporting we have cited it has rarely been used, roughly 250 holders housed across the decade from 2014 to 2024, a figure we report as the city's and not ours ^[10]. Set that beside an estimated 5,893 displaced households and the present-day shape of return looks as partial as the rebuild itself. We note it with attribution and without extrapolation. It is the live edge of a sixty-year story, and it is not ours to resolve.

A fuller accounting would require something this paper does not have, and naming what it would require is part of being honest about the gap. To measure whether the rebuild repaired the harm, a researcher would need to follow households, not buildings. That means a record of who was displaced from the Western Addition in the 1950s and 1960s, where they went, what became of them and their children, and whether any of them or their descendants now live in the subsidized units that stand on the cleared blocks. Some of that record exists in fragments, in the Certificate-of-Preference rolls the city keeps, in oral histories, in the relocation files the Redevelopment Agency generated at the time. None of it is in the portfolio we read, and assembling it would be a different project demanding different sources and a different method, closer to the household-tracing work that Kauer and colleagues model than to a read of a public housing portfolio ^[11]. The serial-forced-displacement literature points the same way, toward a long, person-centered record of repeated upheaval on the same ground rather than a snapshot of what currently stands on it ^[5]. We flag the shape of that fuller accounting to mark precisely where our evidence stops, so a reader does not mistake the edge of our data for the edge of the story.

Two last honesties before the page closes. The snapshots will drift, because both files are point-in-time reads of a portfolio the city keeps updating, so the counts here are true to the data as we read it and not to a future that will revise them ^[1]. The 2020s figure in particular carries future-dated and in-progress construction, the latest date in the file falling in 2027, so the most recent stretch of the rebuild has not finished arriving and will resolve over time ^[1]. Better to flag that the late numbers are soft than to let them read as settled.

Forty years of public subsidy on cleared ground describes a partial and slow response set against a fast and near-total removal. That sentence is the most the evidence permits. The removal happened in roughly a decade and a half and took most of a district. The rebuild, in the record we can read, has taken four decades and replaced a small fraction of what stood. We have set the two beside each other at their real sizes and gone no further, because going further would mean inventing the people and the causes this portfolio cannot hold. A fuller accounting exists, but not in this dataset. It lives with the displaced households themselves, and in the records of repeated upheaval that the serial-forced-displacement literature has begun to gather, the redlining and renewal and disinvestment that fell on the same blocks in sequence ^[5]. Until that accounting is made, the cleared ground tells us what was rebuilt and withholds who was lost. The most faithful thing we can do is say exactly that much, and not a word more.