Some of the numbers are the best in years. Some are the worst on record. We analyzed every public dataset San Francisco maintains to sort out which is which.

1I. The Paradox

Crime is down 45 percent. Overdose deaths peaked and are falling. The Board of Supervisors is less contentious than it has been in years. Yet the last City Survey — conducted in 2023, before much of the improvement documented here — found San Franciscans rating their satisfaction at the lowest level in twenty-seven years of polling. The turnaround is real, at least in parts. The misery is also real. Both are happening at the same time, in the same forty-nine square miles, and neither one is lying.

San Francisco is seven miles by seven miles. You can walk across it in an afternoon. It packs contradictions into those square miles the way it packs Victorians onto its hillsides: a studio apartment rents for $2,800 a month and a man dies of an overdose on the sidewalk outside. The most progressive major city in the country recalled its school board, recalled its district attorney, and elected a mayor who had never held public office. The birthplace of the United Nations, the Summer of Love, and the modern LGBTQ rights movement is also the place where open drug use and tourist selfies coexist on the same block.

Every San Franciscan carries their own version of this city. The version where the tech money ruined it. The version where the progressives ruined it. The version where nobody ruined it and the fog just makes everything look dramatic. None of them are entirely wrong.

But something shifted in the past five years. The contradictions stopped feeling productive and started feeling pathological. The national narrative — amplified by cable news, by tech moguls with Twitter accounts, by politicians who had never ridden the 38 Geary in their lives — hardened into a simple story: San Francisco is a failed city. A cautionary tale. A place that proves progressivism doesn't work.

That narrative is wrong. But so is the counter-narrative — the one you hear from boosters and mayors and tourism boards — that everything is fine, that the city is "back," that the vibes have shifted. Both stories are too simple for a place this complicated.

Almost nobody has looked at the data — not the cherry-picked numbers that prove a point, but the full, contradictory record of what actually happened. This article is the result: 19.5 million rows of city data, 1,310 meeting transcripts discourse-analyzed by NLP, 117 years of ballot measures, and 27 years of resident surveys.

What follows is complicated. The city is genuinely improving in some areas and deteriorating in others, and in a few cases the ground has shifted so fundamentally that the old benchmarks no longer apply.

2II. A City That Governs by Argument

To understand where San Francisco is now, you have to understand what it has always been: a place where the argument is the governance.

Since 1907, San Franciscans have voted on 1,998 ballot propositions. One thousand, nine hundred and ninety-eight. They have voted on whether to fluoridate the water (yes, 1951), whether to name the sewage treatment plant after George W. Bush (no, 2008, though 31% said yes), and whether to tax sugary beverages (yes, 2016). They have voted on rent control, on stadium deals, on homeless encampment policies, on Uber and Lyft, on affirmative action. Of these 1,998 propositions, 1,203 passed, 763 failed, and 32 were withdrawn or rendered moot before a final vote. The city does not merely elect representatives and let them govern. It second-guesses, overrides, instructs, constrains, and occasionally contradicts them — all through the ballot box, twice a year, with the regularity of the tides.

This is not a bug. It is the operating system. San Francisco invented itself through direct democracy, and every attempt to understand its politics without understanding this fact will fail. The Board of Supervisors passes 77.6% of the legislation it considers — 7,214 out of 9,293 items since 2019. Only thirteen bills have outright failed. Three have been vetoed. The Board, in practice, functions as a near-unanimous ratification body. The real fights happen elsewhere: at the ballot box, in the commissions, at the school board, and — most of all — in the street-level reality that no meeting transcript captures.

The tension that has defined the city since at least the 1970s is between two strains of left-of-center politics that, in most American cities, would be allies. The first is the progressive tradition — tenant protections, social spending, skepticism of development, solidarity with marginalized communities. The second is the pragmatic tradition — build more housing, fund the police, run the trains on time, don't let the perfect be the enemy of the functioning. In San Francisco, these two camps do not argue across a partisan divide. They argue across a dining table. They both voted for Biden by 85 points. They both recycle. They both have strong opinions about the Muni.

This argument — progressive versus pragmatic, protection versus production, compassion versus competence — has been the city's creative engine and its central dysfunction for half a century. And every crisis forces it back into the open.

By 2019, the argument had been running hot for years, fueled by a tech boom that was transforming the city's economy and demographics at a pace not seen since the Gold Rush. The numbers that year told a story of a city at the peak of its material prosperity — and already cracking under the weight of it.

There were 119,177 reported crime incidents in 2019. BART carried 118 million riders across the Bay. San Francisco International Airport moved 57.4 million passengers. The city's building permit pipeline contained 36,832 net new housing units in various stages of filing and approval. Sales tax revenue hit $124 million. Nineteen thousand four hundred and thirty-one new businesses registered. The San Francisco-Oakland-Hayward MSA unemployment rate was 2.7% — effectively full employment. The city budget was $11.1 billion, a number that would have been inconceivable a decade earlier.

But prosperity was not evenly distributed, and it was not gentle. Eviction filings were running at 1,456 per year. The Department of Homelessness and Supportive Housing was spending $453 million annually — already more than the entire city budget of many mid-sized American cities — and the tent encampments along Division Street, under the freeway overpasses near SoMa, were growing, not shrinking. The Tenderloin, which has been the city's open wound for as long as anyone alive can remember, was deteriorating block by block: more needles, more foil, more people sleeping in doorways, more ambulance calls. Residents on Turk and Eddy and Leavenworth streets were calling 311 so often that the line was functionally a neighborhood hotline. Across the city that year, 311 received 657,000 calls.

And on Mission Street, the Google buses — those sleek, unmarked shuttles that carried tech workers from their San Francisco apartments to their Silicon Valley campuses, a forty-mile commute through a world they passed through without touching — had become such a potent symbol of displacement that activists had blockaded them and turned them into the defining image of a city at war with itself. The Google buses have largely faded, diminished by remote work and Zoom calls. But the tensions they embodied — who the city is for, who gets to stay, who decides — never left.

3III. The Pandemic and Its Aftershocks

The Shock

San Francisco was one of the first cities in America to shut down, and it shut down hard. On March 16, 2020, Mayor London Breed and the health officers of six Bay Area counties issued a joint shelter-in-place order — three days before California's statewide mandate, a full week before New York. For a brief moment, the city's willingness to act early and aggressively looked like the best kind of San Francisco: progressive, science-driven, decisive. The early COVID death toll was lower than almost any comparable American city. The economic cost was something else entirely.

San Francisco International Airport, which had moved 57.4 million passengers in 2019, dropped to 16.4 million — a 71% collapse. BART ridership cratered. The system had carried 118 million riders in 2019; by the first full pandemic fiscal year, it had lost nearly three-quarters of them. The stations at Embarcadero, Montgomery, Powell, and Civic Center — the four stops that define the downtown corridor, the stations that made commuter San Francisco possible — turned into echoing concrete galleries where the loudest sound was the escalator.

Employment across the MSA fell 9.2%. Unemployment, which had been 2.7% before the pandemic, spiked to 13.9% in April 2020. More than four hundred thousand jobs disappeared from the San Francisco-Oakland-Hayward metro area at the pandemic’s trough. New business registrations dropped 25%. Sales tax revenue — the most direct measure of commercial vitality — cratered roughly 40% from its 2019 baseline.

And then something happened that is still not fully understood, and that explains more about San Francisco's current predicament than any other single fact: the city budget did not shrink. It grew. Federal emergency dollars — CARES Act, American Rescue Plan, FEMA reimbursements — flooded in, and the budget increased 11% in FY2020-21, to $12.3 billion. The city was hemorrhaging private-sector revenue and replacing it with federal money on a one-for-one basis, and then some.

This was not unique to San Francisco. Every major American city received emergency federal funding. But San Francisco's particular version of this phenomenon — a city whose tax base was collapsing while its government was expanding — set the template for what came next. Because the money was there, there was no forcing function for reform. Because there was no forcing function for reform, every existing program continued, every existing department maintained headcount, and the crisis created an opportunity to dramatically expand spending on homelessness without having to cut anything else.

Homelessness spending nearly doubled in a single fiscal year, from $453 million in FY2019 to $844 million in FY2021. Much of this went to Project Roomkey, the program that placed homeless individuals in hotel rooms — a necessary emergency intervention that became, in the eyes of many residents, a symbol of the city's tendency to spend enormous sums without measurable results. As a state audit later found, California broadly failed to track its homelessness spending outcomes.

The money saved lives. The question of whether it changed outcomes — whether the trajectory of homelessness in San Francisco was meaningfully altered by spending $844 million instead of $453 million — is one the city has never convincingly answered.

The Schools War

If the pandemic revealed the best of San Francisco's crisis response in March 2020, the school board revealed the worst of its governance in February 2021.

San Francisco's public schools were among the last major urban districts in the nation to reopen for in-person learning. Parents were frantic. Kids were falling behind. And on January 26, 2021, while schools remained closed, the San Francisco Board of Education voted to rename 44 schools — Abraham Lincoln, George Washington, Dianne Feinstein — using reasoning that ranged from defensible to absurd.

We know exactly how that meeting felt, because we measured it. Our NLP analysis of the February 9, 2021 Board of Education transcript — a meeting dominated by the Lowell High School admissions overhaul — scored it at a contentiousness level of 78 out of 100 — the single highest contentiousness score in our entire dataset of 1,310 analyzed transcripts. Seventy-eight. To put that in perspective, the average Board of Supervisors meeting today scores a 19 (the historical average is 34). The average city meeting across all bodies scores a 21. This was not a contentious meeting. It was the most contentious public meeting in San Francisco in the period we studied.

The March 2021 school board meeting — the one that finally, agonizingly, addressed the question of when children might actually return to their classrooms — scored 72.

The school board is, by our data, the most contentious public body in San Francisco. Its average contentiousness score of 53 dwarfs the Board of Supervisors' 34. More striking still: public engagement at school board meetings averages 66.4 on our scale, compared to 27.9 at the Board of Supervisors. Parents show up. They testify. They shout. They cry. They bring their children. The school board, not the Board of Supervisors, is where the real democratic intensity of San Francisco concentrates — and it is, not coincidentally, where the city's governance failures are most visible and most painful.

In February 2022, San Francisco voters recalled three school board members — Alison Collins, Gabriela Lopez, and Faauuga Moliga — by margins between 72 and 79 percent. It was the first successful school board recall in San Francisco's history. The new board, appointed by Mayor Breed, has faced its own turbulence — SFUSD enrollment continues to decline, the district confronted a $113 million budget deficit in 2024, and school closures became the defining controversy of the post-recall era. The recall changed the personnel. Whether it changed the outcomes remains an open question. It was not the last recall the city would see that year.

The Boudin Recall

Chesa Boudin was elected District Attorney in November 2019 on a platform of progressive criminal justice reform. He was the son of Weather Underground radicals, raised by Bill Ayers and Bernardine Dohrn, a former deputy public defender who had campaigned on restorative justice. He won a four-way ranked-choice race with 50.8% of the final-round vote, inheriting an office that processes roughly 25,000 cases a year. As Mission Local reported on the night of his recall, the election was as much a verdict on the city's direction as on Boudin himself.

His timing was catastrophic. Within four months of taking office, the pandemic hit. The courts shut down. The jails began releasing inmates to prevent COVID outbreaks. The police, already skeptical of a DA who had campaigned on holding officers accountable, pulled back. And the drug market that had been slowly consuming the Tenderloin and SoMa accelerated into something no one had seen before.

The numbers are stark: drug offense reports dropped 62% during Boudin's tenure, from 2,292 in 2019 to 873 in 2021. Whether this represents a deliberate deprioritization by the DA's office, a collapse in police referrals, a change in reporting practices, or some combination of all three is still debated. What is not debated is what anyone with eyes could see: the open-air drug market in the Tenderloin expanded from a few blocks to an entire neighborhood. People were smoking fentanyl in broad daylight. The sidewalks on some blocks became impassable. The smell of human waste was inescapable.

The national media arrived and amplified what was already visible to anyone walking through the Tenderloin. Camera crews from Fox News and CNN alike set up at the corner of Hyde and Turk. The footage was undeniable. The question was what it meant — whether these blocks represented all of San Francisco or a concentrated crisis in a city of forty-seven square miles.

Boudin was recalled on June 7, 2022, with 55% of the vote. The recall was held concurrently with the June statewide primary, which saw turnout of roughly 46 percent of registered voters.

The recall defied easy national framing. San Francisco had voted 85% for Joe Biden. It continued to elect progressive supervisors. But it had also just voted, by double digits, to remove a progressive DA whose policies many residents blamed for the disorder on their streets. Progressives argued the recall was funded by billionaires and scapegoated a DA who inherited a broken system. Recall supporters argued they were holding an elected official accountable for visible, measurable failure. The data supports elements of both interpretations — and the gap between what San Francisco said it believed and what it was willing to tolerate had become untenable.

The Fentanyl Crisis

The recalls — of the school board, of the DA — were political earthquakes. But the crisis that truly changed the city was not political at all. It was pharmaceutical. And it arrived on foil.

Fentanyl had been present in San Francisco's drug supply since at least 2015, but the wave that hit in 2020 was different in kind, not just degree. The drug was cheaper, stronger, and more lethal than heroin. It arrived in a city whose social safety infrastructure was already overwhelmed, whose emergency rooms were already full, and whose streets — emptied by the pandemic lockdown — had become, in effect, an open-air hospice.

The overdose death count tells the story with terrible clarity: 725 people died in 2020. Then 641 in 2021 — a brief, deceptive dip that some mistook for progress. Then 647 in 2022. And then 2023, the worst year, when 810 people died of drug overdoses in San Francisco. That is 2.2 people per day, every day, for an entire year. In April 2023, 71 people died. In August, 88 — the worst single month in the dataset.

To put those numbers in human scale: the fentanyl crisis killed people at roughly six times the rate of homicides. San Francisco averages about 50 murders a year. It was losing that many people to overdoses in a single month.

This was the crisis that changed everything — not because it was new, but because it was visible. The encampments, the needles, the foil, the blue lips of an overdose — it happened on the sidewalk, in front of City Hall, in front of the Asian Art Museum, in front of the children walking to school.

But the Tenderloin is also a neighborhood where people live. Families walk their kids to Tenderloin Elementary. Seniors line up at Glide for meals served there for sixty years. The crisis is real. The people who live inside it deserve more than someone else’s narrative.

And the data captures the institutional vacuum that accompanied it. Drug offense reports had bottomed at 873 in 2021 — that 62% collapse from 2019. The gap between the scale of the crisis and the scale of the enforcement response was staggering.

The Mood of City Hall

While the city was living through these overlapping catastrophes, its government was meeting, debating, passing resolutions, and — we now know, because we measured it — experiencing a seismic shift in its own emotional register.

Our NLP analysis of 1,310 meeting transcripts captured the shift in real time: urgency scores jumped 43% during COVID, contentiousness barely moved — and then the real divisions came. The pandemic united San Francisco's government. The aftermath divided it. The full analysis follows in Section VI.

4IV. London Breed's San Francisco

London Breed became mayor of San Francisco in July 2018, winning a special election to replace Ed Lee, who had died in office in December 2017. Breed had briefly served as acting mayor by virtue of being Board President, before the Board voted to install Mark Farrell as interim caretaker — a move that prevented Breed from building incumbency advantage. She won the June 2018 election anyway. She was the city's first Black woman mayor, raised in public housing in the Western Addition, a graduate of Galileo High School and UC Davis. She had served on the Board of Supervisors, been its president, navigated the city's factional politics with a skill that reflected years in the city's political trenches. She was 43 years old. She had no idea what was coming.

For all the attention paid to San Francisco’s progressive politics, the city has not had a mayor from the progressive wing since George Moscone was assassinated in 1978. Breed, like Brown, Newsom, Lee, and Farrell before her, governed from the pragmatic center. The progressive tradition that shaped the Board of Supervisors never controlled Room 200.

No mayor could have been prepared for what San Francisco experienced between 2018 and 2024. Breed's tenure spanned the entire arc of crisis and partial recovery: the late boom, the pandemic, the shutdowns, the recalls, the fentanyl emergency, the national narrative of failure, and the uncertain beginning of whatever comes next. To evaluate her record honestly — neither the hit piece her opponents want nor the redemption arc her supporters prefer — requires holding multiple truths simultaneously. It requires, in other words, thinking like a San Franciscan.

She spent more on homelessness than any mayor in the city's history. The Department of Homelessness and Supportive Housing budget rose from $453 million in FY2019 to $835 million by FY2025 — an 84% increase during her tenure, and part of a longer arc that had taken the department from $217 million in FY2017 to nearly $700 million projected for FY2027. In a wide-ranging exit interview with KQED, Breed reflected on these investments and the frustrations that accompanied them. She opened shelter beds, expanded the Homeward Bound bus ticket program, created the Street Crisis Response Teams. The spending was real. The results were — and this is the most painful sentence a journalist can write about a public official who appeared to be genuinely trying — not measurable. The encampment reports to 311, which totaled 140,409 over the period, did not decline. The overdose deaths did not decline. The Our City, Our Home Oversight Committee, created by voters to monitor the spending, registered the highest urgency scores of any non-education body in our NLP analysis.

She navigated the pandemic response with a decisiveness that earned national praise. San Francisco's early lockdown, issued March 16, 2020, was credited with saving lives. Her administration distributed millions in emergency aid to small businesses and tenants. But the schools stayed closed — longer than in Los Angeles, longer than in New York, longer than in most comparable cities — and the political cost of that decision, however constrained by public health officials and union negotiations, was severe. As enrollment continued to plummet in the aftermath, some schools faced outright closure.

She declared the fentanyl emergency. She pushed for conservatorship legislation. She deployed police to the Tenderloin in concentrated operations. She stood at press conferences and said the things that needed saying: that the open drug markets were unacceptable, that compassion without accountability was not compassion, that people were dying on the streets of a city that spent more on social services than most states.

And some things got better. Crime began declining significantly on her watch: total incidents dropped 18.7% between 2023 and 2024, with notable reductions in property crime. The overdose death count fell from its catastrophic peak of 810 in 2023 to 635 in 2024, then to 622 in 2025 — still terrible, still two people a day, but trending in the right direction for the first time in years. SFO airport recovered to approximately 91% of its pre-pandemic passenger volume. Sales tax revenue clawed back to $105 million, about 85% of the 2019 peak.

But the structural numbers tell a harder story. Housing production collapsed on her watch — building permit applications fell 41% from the 2019 peak and never recovered, while housing completions dropped 74%. Market-rate construction, the engine that had driven the city's housing pipeline for a decade, essentially stopped. The causes were largely beyond any mayor's control — interest rates, construction costs that rose roughly 40% between 2019 and 2025, a post-pandemic capital market that would not finance new apartment buildings in a city where commercial vacancy had tripled from 6.4% to 23.4%. But the effect was the same: the city was not building, and every year it did not build, the housing crisis deepened and the affordability crisis worsened.

BART ridership never recovered. The system was carrying roughly 43% of its pre-pandemic volume — a permanent structural shift driven by remote work that no mayor, no transit agency, no amount of marketing could reverse. The downtown office corridor, which depended on BART commuters for its economic vitality, was entering what commercial real estate analysts were calling a "doom loop": fewer workers, less foot traffic, more vacancies, lower tax revenue, fewer services, fewer workers. The Bloomberg newsletter pushed back on the narrative, arguing the doom loop framing ignored signs of life — but the structural reality on the ground was hard to dismiss.

And the city budget kept growing. From $11.1 billion when Breed took office to $15.9 billion by FY2025 — a 44% increase over six years. The workforce, meanwhile, shrank 4.5%. The gap was filled by overtime. Police overtime alone totaled $1.61 billion across the dataset period. Across five major departments — Police, MTA, Fire, Public Health, and the Sheriff — overtime spending totaled $5.6 billion. It was not a staffing gap being temporarily bridged. Overtime had become a compensation structure, a parallel payroll system that rewarded chronic understaffing and resisted every reform effort.

London Breed lost her re-election bid in November 2024 to Daniel Lurie, the heir to the Levi Strauss fortune who had founded the nonprofit Tipping Point Community. She ran on her record. Lurie ran on change. The voters chose change — not overwhelmingly, not angrily, but with the weary pragmatism of a city that had been through too much to be satisfied with "better than it was."

The data does not render a simple verdict on the Breed era, because the Breed era was not simple. She governed during the worst public health crisis in a century, a drug epidemic that killed more people than homicides and traffic accidents combined, a fiscal environment warped by federal emergency spending, and a national media narrative that treated her city as a cautionary tale. Some of what went wrong was her fault. Some of it was the fault of a political system that distributes power so widely that no single official can be held accountable. And some of it — the interest rate environment, the pandemic, the fentanyl supply chain, the structural shift to remote work — was the weather. Nobody controls the weather. But you are still the one standing in the rain.

5V. What the Data Actually Shows

The tension at the center of San Francisco right now is between the statistics and the stories. Crime is down 45 percent — but someone’s car was still broken into last week. Unemployment is manageable — but a favorite restaurant just closed. The numbers and the lived experience regard each other across an unbridgeable distance.

What follows is an accounting — every available public dataset the city maintains, cleaned, cross-referenced, and run fresh.

A. Public Safety: The Perception Gap

Here is the headline number: total reported crime in San Francisco fell 35.6 percent between 2019 and 2024, from 119,177 unique incidents to 76,715. In 2025, the decline extended further — 65,707 incidents, a 44.9 percent drop from the pre-pandemic baseline, averaging roughly 5,475 incidents per month across the full twelve months of data.

That is not a rounding error. That is not a reporting artifact. Every single one of the city's ten police districts recorded a decline. Central district, covering Chinatown and the Financial District, fell 46.9 percent. The Richmond dropped 46.6 percent. Northern, which includes Pacific Heights and the Marina, came down 43.4 percent. Park, stretching across the Haight and the Inner Sunset, declined 42.0 percent.

Crime is down by more than a third, and the trend is accelerating. Why, then, did San Franciscans' satisfaction with public safety hit an all-time low of 2.98 in the 2023 city survey — a score below even the anxious post-dot-com-bust nadir of 2002, when the economy was cratering and the Hunters Point Shipyard cleanup was dominating headlines? The answer lies in the composition of what's changing and what isn't.

Larceny theft — the broad category encompassing car break-ins, shoplifting, and package theft — drives approximately 60 percent of the total decline, falling from 48,549 incidents to 23,630, a 51.3 percent drop. This is the category most susceptible to changes in reporting behavior. When you've had your car window smashed for the third time and you know no one will dust for prints, you stop calling it in. The SFPD itself has acknowledged the underreporting problem. As the SF Chronicle reported, policy changes have played a role, but experts caution that the full picture is messier than the headline suggests. The decline in larceny theft is real, but its magnitude almost certainly overstates the improvement in lived experience.

Violent crime, meanwhile, fell 26.6 percent — meaningful but more modest. The composite dropped from 12,105 incidents in 2019 to 8,890 in 2024. Robbery, the street crime most likely to generate fear, dropped 41.3 percent, from 3,374 to 1,981. But assault — the category that includes everything from bar fights to domestic violence to random attacks on the street — proved stubbornly resistant, declining only 19 percent, from 8,148 to 6,594. Assault is the most embodied crime category. It is the one you feel in your bones, the one that shapes whether you walk home a different way. And it barely budged.

Then there is the counter-trend that explains perhaps more than any single number why San Franciscans feel less safe in a city with less crime.

Drug offenses are the only category that increased, and they increased dramatically: from a low of 873 in 2021 to 3,633 in 2025 — a 75 percent year-over-year jump in the most recent data. This reflects an enforcement surge, not necessarily more drug use. When the state sent California Highway Patrol officers and National Guard support to the Tenderloin in 2023, when it expanded open-air drug market interdiction, when Mayor Lurie made visible enforcement a centerpiece of his first months — the arrest numbers followed. But the drug crisis is about more than arrests. It is about death. Fentanyl overdoses peaked at 810 in 2023 — more than two San Franciscans per day, a number that dwarfed the city's COVID fatality rate for that same year. The trend since then has been encouraging but incomplete: 622 deaths in 2025, a 23 percent decline from the peak. The second halves of both 2024 and 2025 averaged approximately 43 deaths per month, down from 67 per month at the crisis's worst. That improvement is real. It is also a pace that would have been considered an unfathomable catastrophe in any year before 2020.

The Tenderloin, as always, occupies its own category. It recorded the smallest crime decline of any police district — just 21.8 percent — and absorbed a disproportionate share of drug offense arrests. The neighborhood functions as a structural exception to every citywide trend, a six-block radius that operates under different gravitational laws. The crime map below shows the pattern district by district.

One final geographic note: the Southern District — covering SoMa, the area around Moscone Center, and the blocks south of Market — was the only district to see crime increase in 2025, reaching 10,354 incidents, above its 2024 level. The timing is suggestive. As enforcement pressure intensified in the Tenderloin, activity appears to have migrated south and east, following the path of least resistance into SoMa's wide sidewalks and vacant storefronts. Displacement, not reduction. It is the oldest pattern in urban policing.

But citywide numbers mask geographic reality. Some districts saw crime cut nearly in half; the Tenderloin barely budged. Hit play to see the geography of the decline unfold.

The prosecution pipeline complicates the story further. Of 118,918 cases that entered the DA’s system during our study period, 41,482 resulted in conviction — roughly 35 percent. Another 14,767 were diverted. And 13,236 were dismissed outright. These numbers span both Boudin’s tenure and Brooke Jenkins’s. The funnel is leakier than either side of that debate acknowledges.

The perception gap is not irrational. San Franciscans are responding to a city where visible disorder intensified even as aggregate metrics declined — where assault barely dropped while car break-ins fell by half. They are weighting what they see on their walk to work more heavily than what appears in a police database.

The City Performance Survey, conducted since 1996, quantifies this gap across twenty-seven years and 42,633 weighted responses. General satisfaction peaked at 3.56 in 2015, slid to 3.37 by 2019, and by the most recent survey in 2023 had fallen to 2.98 on a five-point scale — the lowest rating in the survey’s history, below the dot-com bust of 2002 and the depths of the 2009 recession.

The explanations compound. Visible disorder concentrated in corridors that function as the city’s public face — the Tenderloin, Mid-Market, Civic Center — even as citywide averages improved. The fentanyl crisis transformed street-level drug use into a lethal spectacle. The downtown hollowing left BART at 43% and Market Street storefronts shuttered. San Francisco became a national political punchline — coverage that was often exaggerated, sometimes dishonest, and not entirely wrong. And the people who would have given higher ratings left: the city lost 66,000 residents between 2019 and 2023, disproportionately the young professionals who drove the 2015 satisfaction peak. The 2023 survey is not measuring the same city.

But that survey is also a snapshot taken before much of what this article documents. No survey has been conducted since. Crime has continued to fall — down another 14% in 2024 and 10% more in 2025. Overdose deaths have declined 23% from their peak. The political system has realigned. Whether resident sentiment has followed those trendlines, or whether the perception gap has become self-sustaining, is a question the discontinued survey can no longer answer.

B. Housing: A Different Crisis Than Everyone Thinks

The San Francisco housing conversation has been stuck in the same groove for so long that most people don't realize the record has changed. For twenty years, the dominant narrative was about cost — about median home prices pushing past a million, about tech workers bidding up rents, about teachers and nurses and firefighters who couldn't afford to live in the city they served. That story was true. It is still true. But it is no longer the most important housing story in San Francisco.

The most important housing story is that the city has nearly stopped building.

Building permit applications — the earliest indicator of future construction — crashed 41 percent in the first pandemic year of 2020, from 39,236 to 23,038, and never recovered. Volume stabilized at roughly 23,000 to 24,500 per year from 2021 through 2025 — about 62 percent of the pre-pandemic baseline. That stabilization was a floor, not a recovery.

But permits are applications. What matters is what gets built. Housing completions peaked at roughly 5,000 units in 2020, which sounds like a pandemic-era bright spot until you realize those units were the tail end of the pre-pandemic pipeline — projects approved in 2017 and 2018 finally finishing construction. By 2024, completions had fallen to approximately 1,700 units, a 66 percent decline from that peak.

Now separate those completions into their two component parts, and the picture sharpens into something genuinely alarming.

The private market has essentially stopped delivering new housing in San Francisco.

Market-rate completions — the housing produced by the private sector in response to market signals — fell from roughly 4,000 units at the pipeline’s peak to 695 in 2024. An 83 percent collapse. That is the single most important housing number in this analysis. It means the private market has essentially stopped delivering new housing in San Francisco. High interest rates are part of the explanation. Construction costs that have risen roughly 40 percent since 2019 are part of it. An approvals process that remains among the most onerous in the country is part of it. But the cumulative effect is a city that has gone from producing thousands of new market-rate homes per year to producing fewer than seven hundred.

Affordable housing completions also declined, from 1,342 units in 2021 to 683 in 2024. But here the math produces an irony sharp enough to cut: because market-rate production collapsed so much faster, the affordable share of total completions actually increased — to roughly 50 percent of all units delivered. That is not an achievement. It is the arithmetic consequence of a private-sector construction market in free fall. San Francisco is not building more affordable housing. It is building so little of everything else that the affordable pipeline — sustained by public subsidy and operating on its own slower timeline — now constitutes half the output.

The state's Regional Housing Needs Allocation — the RHNA targets that Sacramento imposes on every California jurisdiction — quantifies the gap. For the current cycle, San Francisco is expected to permit 20,518 units in its first two years. It permitted 4,559. That is 22.2 percent of target. Not 22.2 percent of some aspirational stretch goal. Twenty-two percent of what state housing analysts determined was the minimum necessary to address the region's housing shortage. Community leaders have warned that the city's housing promises are unlikely to materialize at the pace required.

Even the most celebrated tool of small-scale housing production — the accessory dwelling unit, the in-law apartment, the backyard cottage — has cratered. ADU permits fell from 400 in 2020, when pandemic-era flexibility and state preemption laws created a brief golden window, to 85 in 2024. A 79 percent decline in the housing type that was supposed to be the easiest, fastest, most politically palatable path to more supply.

But if the production numbers are grim, the eviction data tells a story that is more consequential — and almost entirely absent from the public conversation.

Total eviction notices filed in San Francisco have remained roughly flat: 1,456 in 2019, 1,494 in 2025, a 2.6 percent increase that on its face suggests stability. Below the surface, the character of eviction in San Francisco has undergone a fundamental transformation.

In 2019, non-payment evictions — the kind where a tenant simply cannot pay rent — accounted for just 76 filings. Five percent of the total. The dominant categories were breach of lease (488 filings, 34 percent), owner move-in (198 filings, 14 percent), and Ellis Act withdrawals (116 filings, 8 percent). This was what eviction looked like in the old San Francisco housing crisis: landlords invoking legal mechanisms to remove tenants and capture value from the speculative real estate market. It was brutal, it was often racialized, and it generated an entire ecosystem of tenant advocacy and protective legislation.

That crisis has been supplanted by a different one.

By 2025, non-payment evictions had surged to 488 filings — a 542 percent increase. Their share of total evictions rose from 5 percent to 33 percent. They are now the single largest eviction category. Meanwhile, Ellis Act filings fell 54 percent, to 53. Owner move-in dropped 51 percent, to 97. Breach declined 39 percent, to 300.

Read those numbers together and they describe a tectonic shift. The old eviction crisis was speculative displacement — investors and landlords pushing tenants out to capture rising property values. The new eviction crisis is poverty-driven non-payment — tenants who cannot make rent in a city where real wages have fallen, where the pandemic shattered informal employment networks, where the end of emergency rental assistance programs left thousands without a safety net. The first crisis demanded tenant protections and rent control enforcement. The second demands income support, employment programs, and a social safety net that San Francisco's existing infrastructure was never designed to provide. The policy toolkit from one era does not address the other.

The geographic concentration makes the transformation even starker. District 6 — the supervisorial district covering the Tenderloin and much of SoMa — recorded 46 eviction notices in 2019 and 413 in 2025. An increase of 798 percent. More than any other single data point in this analysis, that number captures what has happened in the blocks south of Market Street: a neighborhood that was already fragile, already over-indexed for supportive housing and social services, absorbed wave after wave of pandemic-era displacement and emerged as the epicenter of a new kind of housing instability.

The final irony, and the one that should keep planners awake at night: District 6 also accounts for 52.2 percent of all net new housing units filed in the city. The same supervisorial district that absorbs the majority of new construction also generates the most evictions. The city isn’t resolving its housing contradictions so much as concentrating them in six square miles.

C. The Economy: Two Recoveries

The simplest way to understand San Francisco's economic position is that there are two economies, and they recovered at different speeds. One of them has nearly returned to its pre-pandemic state. The other has not, and probably never will.

Start with what came back.

San Francisco International Airport — for decades the city's most reliable economic barometer, the single physical chokepoint through which money and ambition flow — processed an estimated 54.5 million passengers in 2025. That is approximately 95 percent of the 57.4 million who passed through in 2019. The airport is not yet whole, but it is close enough that the remaining gap is largely a function of reduced business travel rather than any broader structural shift. People are coming to San Francisco. They are just coming for different reasons.

Unemployment, that blunt instrument of economic measurement, fell back to 3.0 percent in 2022 — essentially full employment — before drifting upward to 4.4 percent as the tech sector began its correction. By the standards of American cities, and certainly by the standards of San Francisco's own history, 4.4 percent unemployment is unremarkable. It is not a crisis number.

But these surface indicators conceal a deeper story of structural loss.

The San Francisco-Oakland-Hayward metropolitan statistical area — the broadest measure of the regional labor market — remains 3.8 percent below its 2019 employment level. In absolute terms, that is 95,000 jobs that existed before the pandemic and do not exist now. Some were eliminated by automation. Some migrated to other metros — Austin, Miami, the suburban rings of the Bay Area itself. Some were simply extinguished, the small businesses and sole proprietorships that closed during lockdown and never reopened.

Real hourly wages — the number that determines whether working San Franciscans are actually better off — declined approximately 5 percent. Nominal wages rose an impressive 19 percent. But cumulative inflation of roughly 25 percent over the same period consumed the gains and then some. The median San Francisco worker earns more dollars than in 2019 and can buy less with them. This is the universal American story of the post-pandemic economy, but it carries particular weight in a city where the cost of living was already the highest in the nation.

Business registrations provide another lens on the same structural shift. New registrations are running at approximately 78 percent of 2019 levels — roughly 15,148 in 2024 against a pre-pandemic baseline of 19,431. There is no recovery trend in the data. The line is flat. Whatever entrepreneurial energy San Francisco is generating, it is not replacing what was lost.

Sales tax revenue, which tracks consumer spending in the city's physical geography, tells the starkest version of the story. In real terms, sales tax collections have declined approximately 35 percent from 2019. They have been declining even in nominal terms since 2022 — the commercial ecosystem is contracting in both real and nominal terms simultaneously.

But the number that explains everything — the single data point that connects the commercial vacancy crisis, the restaurant closures, the quiet emptiness of once-thriving downtown blocks — is BART ridership.

BART carried 42.8 percent of its 2019 passenger volume in the most recent data. Fewer than half of the people who rode BART before the pandemic are doing so now. The disruption is permanent — a structural transformation of how the San Francisco economy functions. The implications cascade through every sector that depended on the daily tide of 200,000-plus commuters flowing through the Embarcadero, Montgomery, Powell, and Civic Center stations.

This is why commercial vacancy rates have reached levels unprecedented in the modern history of the city. CBRE, the commercial real estate firm, reports downtown office vacancy at approximately 36 percent — more than one in three offices empty. As the SF Examiner reported, downtown was bracing to turn the page after a difficult year. Our own dataset shows 3.4 percent, but that measures registered vacancy at the Assessor’s Office — a different thing entirely. Either number would have been unthinkable a decade ago.

Into this vacuum, the AI industry has arrived with all the ambiguity of a new regime.

OpenAI took over the old Uber headquarters in the Mission District. Anthropic established its presence south of Market. A constellation of smaller AI firms — the ones that generate breathless coverage in the technology press — has scattered across the city, preferring the neighborhoods south of Market and in the Mission over the Financial District towers that housed the previous generation of tech giants. The AI boom is real. It is generating wealth, hiring engineers, filling restaurants in Hayes Valley and coffee shops on Valencia Street. But AI companies operate with smaller headcounts per dollar of valuation, work hybrid, and are not leasing the Class A downtown towers. The AI economy is a supplement to San Francisco’s recovery — a trickle where the city needs a tide.

Whether you experience San Francisco as recovering or declining depends almost entirely on which of these two economies you inhabit. Both are right.

D. The Budget: Spending More, Getting Less

San Francisco spends more money per capita than any major city in the United States except New York and Washington, D.C. — and unlike those cities, it does not manage a public school system of a million students or a federal bureaucracy. It is a city and county of 808,000 people with a budget that has grown from $11.1 billion in fiscal year 2019 to $15.9 billion in fiscal year 2025. That is a 44 percent increase in nominal terms, roughly 16 percent after adjusting for inflation. In real dollars, the city is spending about $1.5 billion more per year than it was before the pandemic.

What is it getting for that money?

Start with the revenue structure. The General Fund — the pot of money the city controls for discretionary spending — grew only 29 percent over the same period, substantially less than the 44 percent growth in total spending. The gap was filled by two sources: enterprise fund revenues (airport, water, power) and one-time federal pandemic relief. San Francisco received more than $800 million in CARES Act and American Rescue Plan Act funding, which flowed through Special Revenue accounts and temporarily inflated the city's spending capacity. That money is exhausted. It is not coming back. The spending it enabled — emergency rental assistance, small business grants, expanded health services — is now a structural hole in the budget. Projected budgets for fiscal years 2026 and 2027 show near-zero real growth in revenue. The gap between what the city spends and what it sustainably earns is closing — not through growth, but through the forced austerity of a revenue base that has shrunk with the commercial tax base. The reckoning that budget analysts have warned about for years is arriving.

Nowhere is the spending-versus-outcomes question more acute than in homelessness.

The city's homelessness budget has grown from $217 million in fiscal year 2017 to $453 million in fiscal year 2019 to $835 million in fiscal year 2025. It has nearly quadrupled in eight years. San Francisco now spends more on homelessness, in absolute dollars, than most American cities spend on their entire municipal budgets. The outcomes are, to put it with restraint, not commensurate with the investment.

The point-in-time homeless count has fluctuated within a narrow band — approximately 7,000 to 8,000 people — for the better part of a decade. A state audit found that California broadly fails to track its homelessness spending and outcomes — and San Francisco is no exception. Encampments have migrated geographically but not diminished in aggregate. The visible suffering on the city's streets — the metric that residents actually use, whatever the official counts say — has not meaningfully improved. The city is spending $835 million a year and the problem is, at best, stable.

But consider the math: a single unit of permanent supportive housing in San Francisco costs $700,000 to $900,000 to build, with operating costs of $25,000 to $40,000 per person per year. Divide $835 million by those unit costs and the figure that sounds enormous starts to look like what it actually costs to house people in the most expensive real estate market in America. Whether the money is enough is a different question than whether it is wasted.

Where does the money go? Overwhelmingly, to a network of nonprofit service providers: HealthRight 360 ($429 million cumulative), Episcopal Community Services ($392 million), the Tenderloin Housing Clinic ($390 million), Five Keys Schools and Programs ($251 million). Well over a billion dollars to four providers alone — with an accounting of outcomes the city has been reluctant to provide. The pattern is a microcosm of a broader shift: the City Grant Program, which channels money to nonprofit contractors, has more than doubled from $758 million to $1.66 billion. Functions that in most cities are performed by municipal employees are in San Francisco contracted to hundreds of nonprofits, each with its own overhead and board. The system creates a constituency for the spending itself — organizations whose institutional survival depends on the continuation of the problem they were created to address.

The city's own workforce has shrunk modestly, from 44,525 employees in 2019 to 42,543 in 2025 — a 4.5 percent reduction. But overtime spending has moved in the opposite direction, rising from $320 million to $483 million, a 51 percent increase. The city is paying fewer people to work more hours at premium rates.

The police department illustrates this dynamic in its most extreme form. SFPD headcount dropped 7 percent, from 3,313 officers to 3,079. Overtime spending rose 140 percent. Per-officer overtime averaged $50,283 in the most recent data — up from $19,446 in 2019, a 159 percent increase. The Sergeant III cohort — 417 employees — averaged $373,000 each in total compensation, including roughly $100,000 in overtime per person.

San Francisco is paying its police officers lavishly to do a job no one else wants to do. It is also burning through its existing force with unsustainable overtime loads — generating costs that compound every year without solving the underlying staffing shortage. Both readings are correct.

The compensation divergence between public and private sector workers adds a political dimension to the fiscal story. City employees have gained approximately 11 percent in real compensation since 2019 — their raises have outpaced inflation. Private-sector workers in the same city have seen real wages decline roughly 5 percent. The cumulative gap is approximately 16 percentage points.

This is the kind of divergence that generates ballot measures. It is the raw material of taxpayer revolts, of Proposition 13-style convulsions that restructure governance for a generation. Whether it tips over into that kind of political event depends on factors beyond the data — leadership, narrative, the presence or absence of a catalyzing moment. But the fuel is there. It is accumulating. And the projected austerity budgets of the coming fiscal years will test whether a city accustomed to spending its way through problems can adapt to a world where the money is no longer there.

6VI. What the Meetings Reveal

San Francisco's boards, commissions, and agencies are required by the Brown Act to conduct their business in public. The result is an archive of transcripts — thousands of hours of elected officials, appointed commissioners, and members of the public deliberating, arguing, pleading, and occasionally agreeing on how to run a city. It is, in raw form, democracy's paper trail. In raw form, it is also almost completely illegible — hundreds of pages per meeting, dozens of meetings per month, years of accumulated institutional language that no human being could read in a lifetime.

So we did something that, to our knowledge, has not been done before. We fed 1,310 San Francisco government meeting transcripts — spanning from 2019 through early 2026, covering every major board, commission, and agency — into a large-language-model analysis pipeline and asked it to measure five dimensions of each meeting's discourse: urgency (how pressing the discussion felt), contentiousness (how much disagreement was present), public engagement (how much the public participated), consensus (how much agreement emerged), and policy complexity (how technically dense the discussion was).

The most dramatic pattern is also the most intuitive. In the first quarter of 2020, as COVID-19 arrived in San Francisco and the city entered lockdown, the average urgency score across all government meetings jumped from 37.7 to 54.0 — a 43 percent increase and the single largest shift in any metric across the entire dataset. In the second quarter of 2020, urgency peaked at 59.0. Commissioners and supervisors were making life-and-death decisions about shelter-in-place orders, hospital capacity, and business closures, and the language of their deliberations reflected it.

What is less intuitive — and more revealing — is what happened to contentiousness during the same period. It barely moved.

The pandemic, for all its trauma, was a consensus event. The city's governing bodies agreed, with remarkable uniformity, on the necessity and general shape of the response. Disagreements were tactical, not existential. The public health apparatus had authority and credibility. Supervisors who disagreed on virtually everything else found common ground in the face of a shared emergency. Consensus scores during the acute pandemic phase remained high. The system, when confronted with an unambiguous external threat, functioned.

Contentiousness peaked not during COVID but in 2023, reaching an average score of 25.5 across all bodies. The pandemic united; the reckoning divided. By 2023, the city was arguing about everything the pandemic had deferred and everything it had created: how to handle the fentanyl crisis, whether to recall the school board members (that had already happened), what to do about downtown vacancy, how aggressively to police the Tenderloin, whether progressive criminal justice policies had failed or been sabotaged. These were not technical disagreements. They were arguments about values, about identity, about what kind of city San Francisco intended to be.

Consensus reached its nadir that same year — 76.8, the lowest in the dataset. By 2025 and into early 2026, it had climbed back to 82.0, the highest recorded. This rebound corresponds to the political realignment that swept through the November 2024 election: a new mayor, a shifted Board of Supervisors, a general exhaustion with factional warfare. Whether this consensus reflects genuine agreement or merely the temporary alignment of a new political majority remains to be seen.

The variation across governing bodies is as revealing as the temporal trends.

The Board of Supervisors, the city's legislative branch, operates with a formal contentiousness score of 34.2 and a public engagement score of 27.9. These are moderate numbers — more contentious than the commissions, less publicly engaged than you might expect from the city's most prominent deliberative body. The reason is structural: the Board's real fights happen in committee, where items are shaped, amended, and occasionally killed before they ever reach the full board for a vote. The dissent rate on final roll-call votes is below 1.5 percent. Fewer than two out of every hundred items generate a "no" vote. This is one of the most consensual legislative bodies in American urban governance — not because San Francisco's supervisors agree on policy, but because the institutional machinery filters disagreement before it becomes visible.

Against that backdrop, individual voting patterns become diagnostic. Supervisor Jackie Fielder, representing District 9, carries a 3.2 percent dissent rate — more than triple the board median. In 2026, her dissent rate has risen to 6.4 percent. She is the most frequent "no" vote on a board that almost never says no. Whether you read that as principled independence or ineffective marginalization depends on your politics, but the data is unambiguous: she stands alone more often than any colleague.

At the other end of the spectrum, Supervisor Rafael Mandelman posted a 0.2 percent dissent rate in 2025 — effectively zero. He has aligned perfectly with the emerging moderate majority. The Board passes approximately 1,000 items per year. The true contested terrain exists upstream of the vote, in the conversations and negotiations that shape what reaches the floor. The transcript analysis captures the shadow of those fights — elevated contentiousness in committee hearings, spikes in public engagement on specific agenda items — even when the final vote is unanimous.

The most contentious governing body in San Francisco, by a significant margin, is not the Board of Supervisors. It is the Board of Education.

The school board recorded an average contentiousness score of 53.0 — more than 50 percent higher than the Board of Supervisors. It also registered the highest public engagement score of any body in the dataset: 66.4 versus the Board of Supervisors' 27.9. The real democratic fights in San Francisco — the meetings where parents line up to speak, where voices break with anger, where the gap between institutional language and lived experience becomes unbridgeable — happen not at City Hall but at 555 Franklin Street.

The single highest contentiousness score in all 1,310 transcripts belongs to a specific meeting: the Board of Education's session of February 9, 2021. Contentiousness: 78 out of 100. That was the night the board voted to rename forty-four schools while they were still closed for in-person learning — the meeting whose consequences were detailed in Section II.

The political consequences were swift and severe. Three school board members were recalled by overwhelming margins the following year, in one of the highest-turnout recall elections in the city's history. The NLP analysis captures the moment of rupture with precision — a contentiousness score that towers above everything else in the dataset like a seismographic spike.

At the opposite end of the spectrum, the most consensual bodies operate in near-total agreement. The Fire Commission averaged a contentiousness score of 9.1. LAFCO, the Local Agency Formation Commission, and the Transbay Joint Powers Authority each registered 10.0. These are the unglamorous corners of municipal governance — bodies that oversee hydrant inspections and jurisdictional boundaries and construction oversight — and they function with the quiet efficiency of institutions that are not being asked to resolve the city's deepest contradictions.

The SFMTA — the Municipal Transportation Agency, which oversees Muni, parking, and bicycle infrastructure — occupies a singular position in the data. It registers the highest policy complexity score of any body: 71.0 out of 100. It carries high urgency: 54.5. And moderate contentiousness: 32.7. This is the profile of a technical agency navigating genuine crises — service cuts, operator shortages, infrastructure deterioration, the Muni reliability problems that have driven San Franciscans to distraction for generations — with a level of policy sophistication that outstrips any other body in the dataset. The SFMTA is, in the NLP analysis's quiet judgment, the city's most intellectually serious governing body.

The topical analysis reveals a different kind of story — the story of what San Francisco's government chose to talk about, and how those choices shifted over time.

In 2019, housing dominated the agenda. It was the subject of more formal hearings, more committee discussions, more public comment periods than any other policy area. By 2023, public safety had displaced it for the first time — reflecting the fentanyl crisis, the open-air drug markets, the perception gap we documented earlier. By 2024 and into 2025, homelessness surged to the top of the agenda, driven by encampment policy, the Supreme Court's Grants Pass decision, and the escalating costs we detailed in the budget section.

These shifts trace the arc of a city's attention — the slow migration of institutional focus from one crisis to the next, always reactive, always trailing the lived reality by six to twelve months. But the most striking finding in the topical analysis is not which topics rose but which ones were discussed despite urgency — or not discussed because of institutional inertia.

Homelessness scores an urgency rating of 54.2 — the second-highest of any topic in the dataset. Yet from 2022 through 2023, it received only one to two formal hearings per year from the Board of Supervisors. The system allocates attention by institutional routine — the calendar of required hearings, the rhythm of budget season, the bottleneck of committee agendas — not by urgency. An issue can be universally acknowledged as a crisis and still receive less deliberative attention than a routine land-use variance. The governance gap is not about will. It is about the structural mismatch between how government allocates its most finite resource — time — and how problems actually demand it.

What do the meetings reveal about whether the turnaround is real? The city’s governing bodies have emerged from the most contentious period in the dataset into the most consensual. The political realignment of 2024 produced a Board of Supervisors that votes together more uniformly than any in recent memory. And the public still shows up — at school board meetings, at planning hearings, at the police commission — with an engagement that the NLP system can measure and that anyone who has attended a San Francisco public meeting can feel.

But consensus, by itself, is not competence. The meetings reveal a government that has learned, painfully and publicly, to stop arguing with itself. Whether it has learned to solve problems is the test that matters now.

7VII. The Political Reset

On the night of November 5, 2024, Daniel Lurie stood before supporters at The Chapel in the Mission and said: “It’s time for us to start making people feel like the city loves them back.”

No grand ideological vision. No sweeping progressive manifesto. No law-and-order jeremiad. Just management. The heir to the Levi Strauss fortune, a man who had never held elected office, who had spent his career running a nonprofit called Tipping Point Community that funneled philanthropic dollars toward poverty reduction, had just defeated a sitting mayor in a city where incumbents rarely lose. He did it not by promising to transform San Francisco but by promising to competently administer it.

This was, in its way, the most radical proposition on the ballot.

To understand what Lurie’s election means, consider what he ran on. He is a Democrat who supports abortion rights, climate policy, and immigrant protections. His target was not San Francisco’s political orientation but its operational competence — the Byzantine commission structure, the chronic understaffing, the $15.9 billion budget that somehow never seems to buy enough functioning city. As NPR reported, the fact that Lurie had never held public office was part of his appeal.

The recalls that preceded him — of the school board in 2021, of Boudin in 2022 — were driven by parents, by small business owners, by the Asian American community in the Sunset and Richmond districts, by people who voted for Barack Obama twice and Bernie Sanders once and were tired of feeling like their city didn't work.

Lurie was the mayoral expression of that same impulse: a demand that the city’s government perform as well as its residents expect. You can believe in affordable housing and still be furious that the city approved only 22.2% of its state-mandated housing target. You can support mental health services and still wonder where the $835 million the city spends on homelessness actually goes. Competence is not an ideology. But in San Francisco, by 2024, it had become an insurgency.

He inherits a government that is, by any operational measure, strained. The city's 4.5% staffing deficit means thousands of positions sit unfilled — in public health, in planning, in the permitting offices that are supposed to be approving the housing the city desperately needs. To compensate, the city spent $483 million in overtime in the most recent fiscal year, a 51% increase from pre-pandemic levels. This is the municipal equivalent of paying time-and-a-half for a skeleton crew: expensive, unsustainable, and a guaranteed path to burnout. The structural deficit projections that budget analysts have been warning about since 2023 are no longer projections. They are arriving.

The Board of Supervisors that Lurie works with is, for the first time in years, broadly aligned with the mayor's office. The 2024 election shifted the Board's ideological center of gravity in ways that our legislative data captures with unusual precision.

Start with the tone. Our natural language processing analysis of Board meeting transcripts — 1,310 meetings scored across five dimensions of political discourse — shows contentiousness dropping to 19.1 in the most recent period, the lowest in the entire dataset. Consensus rose to 82.0, the highest. These are not subjective impressions; they are systematic measurements of how supervisors talk to each other, argue with each other, and find agreement. The Board, put simply, stopped fighting.

The legislative output tells the same story from a different angle. The pass rate for legislation climbed to 83.7% in 2024 — again, the highest we measured. Resolutions and ordinances moved through committee and full Board votes with less friction than at any point in recent memory. For a body that had spent the Breed years locked in procedural combat — using hearings, committee delays, and parliamentary maneuvers to block or slow executive priorities — this was a sea change.

And about those hearings: they dropped from 176 to 84. In the Board of Supervisors, hearings are the primary oversight tool — the mechanism by which supervisors investigate city departments, question mayoral appointees, and shine light on policy failures. A hearing on homelessness spending or police staffing or Muni performance is how the Board exercises its check on executive power. Cutting that number in half means the Board is doing substantially less investigative oversight. Whether that represents constructive deference to a new administration or an abdication of the Board's watchdog role depends entirely on what happens next.

The early 2026 data shows the expected friction. Contentiousness jumped to 45.0 — more than double the 2024 figure — while consensus dropped to 67.6. A functional democracy argues; the question is whether this Board can argue productively, without reverting to the procedural paralysis that defined its predecessor.

To understand how this alignment formed, you have to understand the Board it replaced.

From 2020 through early 2024, the Board of Supervisors operated under a progressive majority that at its peak held seven of eleven seats. Aaron Peskin — the longest-serving supervisor in modern San Francisco history, first elected in 2001, termed out in 2009, returned in 2015, and finally gone for good in January 2025 — was its gravitational center. As chair of the Budget and Finance Committee, Peskin controlled what amounted to a second power center. He used it actively: calling hearings, demanding that department heads testify, and holding items in committee that he opposed on the floor. Dean Preston, the democratic socialist in District 5, governed on an explicit platform of rent control expansion and skepticism toward market-rate housing. YIMBY groups accused him of obstructing production. Preston countered that he had approved nearly 30,000 units during his tenure — a claim challenged in a lawsuit that a judge eventually threw out, calling the dispute "linguistic fencing." Hillary Ronen, in District 9, preceded Jackie Fielder and governed the Mission during its most chaotic stretch — she fenced off the plazas at the 24th Street Mission BART station in 2022 after open-air drug markets overwhelmed the neighborhood. "My back is against the wall," she told the SF Standard. Together with Connie Chan, Shamann Walton, and a rotating cast of allies, they formed a progressive bloc that spent the Breed years in sustained procedural combat with the mayor's office.

The shift happened in two waves. In 2022, Joel Engardio defeated incumbent Gordon Mar in the Sunset, flipping the district back to moderate after Mar's single progressive term, and Breed appointed Matt Dorsey — her former police department spokesperson — to replace Matt Haney after Haney won a state Assembly seat. The progressive majority shrank from seven to roughly five. In November 2024, it collapsed. Danny Sauter replaced Peskin. Bilal Mahmood, a tech entrepreneur, replaced Preston — District 5 going from democratic socialist to moderate in a single election cycle, the largest ideological swing on the Board. The result: a 7-to-4 moderate supermajority, the most lopsided alignment since district elections returned in 2000.

The realignment had financial fuel. Since 2020, groups including TogetherSF, GrowSF, and Neighbors for a Better San Francisco spent more than $30 million on campaigns, ballot measures, and political infrastructure — a sum that, by some analyses, exceeded the combined spending of all Board of Supervisors campaigns from 2000 to 2020. Progressives call this the purchase of a political realignment. Moderates call it the most successful civic engagement campaign in a generation. The spending was real. So were the voters.

The supermajority proved durable but not invulnerable. In September 2025, Joel Engardio became the first supervisor in the city's 169-year Board history to be recalled by voters. District 4 residents, who had rejected his championing of the Great Highway closure by a two-to-one margin even as it passed citywide, voted him out by more than 60 percent. Mayor Lurie appointed Alan Wong to fill the vacancy. The moderate count held at seven — but the recall sent a message the rest of the Board could read: alignment with the mayor is not the same as alignment with your district.

Our voting data makes the contours of this realignment visible at a resolution that media coverage rarely achieves. In 2025, about seventy-five legislative items drew at least one No vote — the contested terrain where ideology shows itself. On those items, the moderate bloc voted together with the cohesion of a parliamentary caucus. Sherrill, Sauter, Mahmood, Dorsey, Mandelman, and Melgar aligned between 89 and 93 percent of the time on contested votes. The progressive minority is fractured by comparison. Fielder and Walton — the two most frequent dissenters — aligned on just 63 percent of contested votes. Fielder and Chan, both nominally progressive, agreed only 35 percent of the time. The progressive side does not function as a bloc. It votes as a scattering of individuals who occasionally find common cause.

The individual supervisors tell the story in miniature. Jackie Fielder, the Mission’s democratic socialist, carries a 3.2% dissent rate — more than triple the Board median. In 2026, it climbed to 6.4%, the highest individual-year rate in our dataset. She has cast the only No vote on fifteen items since January 2025, more solo dissents than every other supervisor combined. Shamann Walton’s arc tells the realignment story most cleanly: his dissent rate rose from 0.6% in 2020 to 2.4% in 2025. He didn’t change. The Board changed around him. The same positions that made him a mainstream progressive in 2020 made him a factional outlier by 2025.

At the other pole, Board President Rafael Mandelman’s 1,930 sponsored legislative items make him the most prolific legislator by a wide margin, and his 0.2% dissent rate reflects near-perfect alignment with the majority he helped build. Connie Chan is the hardest to place: 57% alignment with the moderate core, 35% with Fielder, eleven solo No votes in 2025. A progressive who sometimes votes moderate, a dissenter who sometimes dissents alone.

The fault line runs deepest through housing. In December 2025, the Board voted on the most consequential package of zoning reforms in a generation: the Housing Choice-San Francisco program, the Family Zoning Plan, and a suite of General Plan amendments designed to rezone much of the city's west side for denser residential construction. Every measure passed. Every measure passed over the same four No votes: Fielder, Walton, Chan, and Chyanne Chen. The argument from the left — that upzoning without enforceable affordability guarantees amounts to a subsidy for developers — met the argument from the center that the city cannot simultaneously refuse to build and then wonder why a one-bedroom costs what it costs. The votes were 7-4 and 6-4, item after item. Housing is where San Francisco's ideological divide becomes arithmetic.

And then there is the actor that rarely appears in media coverage of Board politics but dominates the legislative record: the Mayor's office itself. Directly sponsoring approximately 150 legislative items per year, the executive branch's pipeline is the single largest source of legislation before the Board. In San Francisco's strong-mayor system, the Board can amend, delay, or reject — but the agenda is substantially set from Room 200. When the Board and Mayor are aligned, as they are now, the city can move with unusual speed. When they are not, as the 2026 contentiousness numbers hint may be coming, the system can grind to a halt. This is the structural reality beneath the personalities: San Francisco's charter gives the mayor enormous power, and the Board's primary leverage is the power to say no.

There is one more feature of San Francisco governance that deserves attention, because it has no real analogue in most American cities: the charter amendment factory.

In 2024, San Francisco voters considered eleven charter amendments. In 2022, eleven. In 2020, eleven. This is not a coincidence; it is a system. San Francisco governs itself, to a remarkable degree, by ballot measure. Commission structures, budget set-asides, departmental authorities, housing approval processes, police oversight mechanisms — the fundamental architecture of city government gets put before voters every two years with a regularity that would astonish a resident of Chicago or Houston. The 2024 crop was characteristically ambitious. Voters restructured commission appointments, streamlined housing approval processes, and modified the balance of power between the mayor and Board on key administrative functions. Each amendment was the product of months of negotiation, coalition-building, and the peculiarly San Franciscan ritual of ballot argument campaigns so detailed they resemble legal briefs.

This system has virtues: it is radically democratic, giving voters direct say over governance structures that in most cities are buried in administrative code. It has pathologies: it creates a labyrinth of budget set-asides and mandated spending formulas that make coherent fiscal planning nearly impossible. The $835 million homelessness budget is not a single line item that a mayor can redirect; it is a patchwork of voter-mandated programs, departmental allocations, and grant-funded initiatives, each with its own constituency and its own political protections.

Daniel Lurie ran on fixing the machine. The charter amendment factory is one reason the machine is so hard to fix — and also, paradoxically, the mechanism through which it might be fixable. In San Francisco, even the tools of reform are complicated.

8VIII. Is the Turnaround Real?

We began this project with a question that seemed simple enough to ask and has turned out to be extraordinarily difficult to answer. Across 33 datasets, 19.5 million rows, 1,310 meeting transcripts, and twenty-seven years of resident surveys, the answer depends entirely on which numbers you weight and which ones you excuse.

What Is Genuinely Improving

Start with public safety, because that is where the data is most unambiguous and most encouraging.

Crime is down 45% across every police district, in every major category, through every change in leadership. The decline is not a blip — it is among the most significant public safety improvements in any major American city in the post-pandemic period.

The overdose curve has bent — deaths down 23% from the 2023 peak, averaging 43 per month versus 67 at the worst. Six hundred twenty-two dead in a year is still a catastrophe, but the direction has changed for the first time.

And governance has shifted: Board contentiousness at 19.1, consensus at 82.0, pass rate at 83.7%. The machine is running. Whether what it produces proves wise is a separate question.

What Is Getting Worse

Now the other side of the ledger, and this is where the turnaround narrative encounters the shoals.

Housing production has collapsed — permit applications down 38% and stuck there, housing completions down roughly two-thirds, market-rate completions down 83%, RHNA attainment at 22.2%. The development pipeline is emptying.

The displacement story has inverted. Non-payment evictions have surged 542% while speculative evictions declined. The crisis has shifted from greedy landlords to poverty — and the policy tools designed for the former are useless against the latter.

BART ridership has stabilized at 43% — meaning 57% of its demand is gone and not coming back. Sales tax revenue has declined 35% in real terms, measuring not a slowdown but an absolute contraction of commerce within city limits.

Business formation stands at 78% of 2019 levels with no recovery trend. In a city whose mythology is inseparable from entrepreneurship, that is more than an economic indicator.

The city workforce is 4.5% below authorized strength while overtime spending has increased 51%. You cannot run a $16 billion organization on a skeleton crew indefinitely.

What Has Changed Permanently

Some of what San Francisco is experiencing is not a downturn but a structural transformation. The distinction matters enormously.

The downtown commuter model is not coming back. Commercial vacancy at 36% is not cyclical — office buildings designed for the commuter economy cannot be easily converted, and the businesses that served 300,000 daily commuters have lost their customer base. Some will adapt. Many will not. This is millions of square feet of built environment designed for a world that no longer exists.

And the federal fiscal cushion is gone. The $800 million-plus in emergency funding that padded budgets from FY2021 through FY2023 has been spent. The next budget cycle will be the first in which the city confronts its structural imbalance without a federal backstop.

The Interconnected Crises

The trends connect to each other, forming a system of reinforcing pressures that no single policy intervention can break.

Housing production collapses, and with it the permit fees and property tax growth that fund city services. Fewer units means longer waitlists, more non-payment evictions, and more demand on a homelessness system already consuming $835 million annually. The downtown hollowing drains the sales tax base, which means less revenue for services already under strain. BART at 43% means fewer commuters means fewer customers means fewer businesses means less tax revenue means less capacity to invest in the transit that might attract riders back. The loop closes on itself.

The city fills the gaps with overtime — a 51% spending increase that amounts to a holding action, not a strategy. And the federal cushion that absorbed the shocks from FY2021 through FY2023 is gone. The next crisis, San Francisco faces with its own diminished resources.

This is the interconnected reality that no single metric captures and no single policy can address. The turnaround is real in some domains and illusory in others, and the domains where it is illusory are the ones that fund the domains where it is real.

The San Francisco Pattern

There is, however, a longer view. And it is worth taking.

On April 18, 1906, the San Andreas Fault sent a magnitude 7.9 earthquake through San Francisco. The fires that followed destroyed 80% of the city — 28,000 buildings, 500 blocks. Three thousand dead, half the population homeless. San Francisco rebuilt in a decade. Not the same city — a different one, with wider streets, a new civic center, and infrastructure that carried it through the twentieth century. In 1989, the Loma Prieta earthquake damaged the Embarcadero Freeway beyond repair — the double-decker highway that had walled off the waterfront since 1959. The city could have repaired it. Instead, it tore the whole thing down and built the palm-lined boulevard that is today one of the most beloved public spaces in America. The Ferry Building, the Embarcadero streetcar, the views of the Bay Bridge at sunset — all of it exists because an earthquake destroyed something that should never have been built.

The dot-com bust of 2001 emptied South of Market. But the infrastructure the boom had built — the fiber-optic cables, the culture of technical ambition — lay dormant until Twitter, Salesforce, Airbnb, and Uber grew into it. The 2008 financial crisis froze construction; when it thawed, the building boom that followed was both an achievement and the affordability crisis that pushed tens of thousands to the East Bay and beyond.

San Francisco does not bounce back. It becomes something else.

We are in another such transition. The pandemic, the fentanyl crisis, the remote work revolution, the downtown hollowing, the political realignment — these are not temporary disruptions from which the city will return to a 2019 baseline. The 2019 baseline is gone. The 874,000-person city with a bustling Financial District and a booming housing market and BART trains packed at rush hour — that city no longer exists, any more than the city of 1905 existed after April 18, 1906.

What is emerging is harder to see because we are inside it. But the outlines are there. Violent crime has fallen to levels not seen in years. The drug crisis, while still catastrophic, has peaked and begun to recede. SFO is humming. The restaurants that survived are booked on Saturday nights. And the other side of the ledger remains brutal: housing production has no credible recovery plan, the commercial core is more than a third empty, transit runs at less than half capacity, and 622 people died of overdoses last year — progress only relative to the even more terrible year before.

The fog does lift. Stand on Twin Peaks at midday in October, when the marine layer that Karl brings through the Golden Gate every summer morning has burned off, and you can see the entire city spread below you — the Richmond marching to the ocean, the downtown towers catching the sun, the bridges spanning the bay in both directions. It is, in those moments, one of the most beautiful cities on Earth. And it is a city that has survived everything the planet has thrown at it for 175 years.

Is the turnaround real?

The honest answer: the trajectory is. Crime is down 45%. The overdose curve has bent. The most consequential housing reforms in a generation have passed. The government has stopped fighting with itself long enough to act. Whether these gains hold — whether they can outrun the fiscal reckoning, the housing collapse, and the structural loss of the commuter economy — is the question that will define the next five years.

San Francisco has survived earthquakes, busts, epidemics, and its own worst instincts. The worst of this passage may be behind it. What comes next depends on whether a city that has always argued about everything can, for once, build as fast as it talks.