[Data Story] San Francisco’s Hidden Pattern: Correlation Between Disorder, Crime and Administrative Strain

I moved to San Francisco recently and noticed quickly that certain neighborhoods FEEL less stable than others. That sense didn’t come from being on endless Twitter refresh or TikTok poverty porn. It comes from everyday cues like sidewalk conditions, storefront upkeep, visible distress, and the overall energy of the street.

That curiosity led me to a bigger question: what, exactly, is behind that gut feeling? Is it just personal perception, or does the data back it up?

There was only one way to find out if this feeling was unfounded or not? Dog-fooding the analysis using Petavue. I spent an entire weekend pushing all the social and crime indicators of the city, published by the government into a Petavue instance and running correlations. Fun weekend!

Here’s what I uncovered.

San Francisco’s situation is often framed as a list of separate issues: rising crime in some areas, homelessness in others, and an overworked service system trying to keep up. But the data suggests something more integrated: across every major dataset, the same few districts repeatedly surface as the most burdened. These are the places where visible disorder, administrative overload, and serious crime stack on top of each other, every day.

Districts where the Issues Converge

Some neighborhoods don’t just feel more strained, the data shows they carry most of the city’s challenges at once. These areas, which we can think of as Convergence Zones, are where crime, disorder, and service demand all rise together.

Here is how San Francisco is divided into supervisor districts:

A few districts consistently rank highest in violent crime. But what makes them stand out isn’t just the crime rate, it’s the mix of conditions surrounding it. These same districts experience the most visible disorder, the highest number of 311 requests, and the heaviest administrative load.

The gaps between districts are large:

• Violent crime ranges from 2.8% to 22.2%, an eightfold difference.
• Nuisance behavior rises 4.5x between the lowest and highest districts.
• Police interaction is nearly 10x higher in the most strained areas.
• 311 requests vary sharply, from 4x to 18x depending on the category.
• Cleanliness complaints in these districts have over 10 complaints per capita, showing how much ongoing cleanup they require.

The Near-Perfect Correlation of Public Safety and Crime

We wanted to dig deeper into the relationship between public safety nuisances and crime severity. When we looked at the data we found the crime categories to be a long list, so we bucketed them by severity, to the best of our judgement.

Here’s how we have grouped the crime categories into simpler buckets for understanding:

The most striking statistical finding is the relationship between High-Risk Crime and Public Safety Nuisances. The correlation coefficient between these two metrics is 0.99. 

This is not merely a strong relationship; it is virtually a mirror image. Where violent crime goes up, nuisance behavior, loud disputes, drug use, suspicious activity rises almost perfectly in sync, and vice-versa. We need to be mindful that this shows correlation, not causation. But the correlation is astonishingly close, nearly perfect.

This convergence of severe violence and pervasive disorder places an unsustainable burden on these communities and the city services designed to support them. They are neighborhoods where city systems are working overtime just to keep up, facing constant system strain in an environment where everything is happening at once.

Digging Deeper and Finding Patterns

San Francisco’s data makes a simple point clear: the everyday signs of disorder people notice on the street aren’t trivial, un-aesthetic, or harmless. They’re closely tied to where the city sees its most serious crime.

Even the correlations with core elements of visible disorder and high-risk crime are extremely strong:

• High-risk crime and Noise have a correlation of 0.87.
• High-risk crime and Cleanliness have a correlation of 0.85.
• High-risk crime and Encampments have a correlation of 0.79.

Crime and disorder are tightly coupled. They show that when a neighborhood struggles with visible disorder, it’s usually dealing with deeper pressures at the same time. Crime doesn’t emerge in isolation, it grows out of environments where strain is already visible and accumulating.

The struggling neighborhood upkeep might not be the cause of crimes. But it correlates very closely.

The Sleeper Indicator: Administrative Strain

Next up, I wanted to see if there is a correlation between the day‑to‑day operational stress of 311 complaints and the crime reported in a neighbourhood.

I came up with a metric for it, defined and saved it inside Petavue.

Administrative General Services (AGS) Rate is a metric that tracks how often residents ask the city for non‑emergency help through channels like 311.

Districts with the highest AGS Rates — ranging from 31.28% to 76.26% — also show sharp increases across all major crime categories. This makes AGS a useful early indicator of neighborhoods where service demand and public safety stress rise together.

Correlation strength between AGS Rate and:

• High-Risk Crime: 0.78
• Medium-Risk Crime: 0.87
• Public Safety Nuisance: 0.79
• Low-Risk Crime: 0.82

Here are the graphs I could plot for quicker visualization.

• Correlation: High-risk crime and Administrative General Services (AGS) Rate

• Correlation: Medium-risk crime and Administrative General Services (AGS) Rate

• Correlation: Public nuisance and Administrative General Services (AGS) Rate

The correlation indicates that Administrative Strain is a sign that a neighborhood is leaning heavily on basic city systems because underlying issues aren’t being resolved. 

But understanding correlation isn’t enough. To grasp why strain persists, we have to look at what support systems exist, and what’s missing.

The Paradox: More Crisis Response, Less Crisis Prevention

The most counter-intuitive finding in San Francisco's data involves the city's deployed resources. When looking at emergency facilities namely, Police Department and Fire Department stations, the city has an appropriate placement strategy. Emergency teams are stationed where incidents occur most frequently.

• Police Presence: High-Risk Crime Districts house 38.1% of all Police Department facilities (8 out of 21).‍
• Fire Presence: These districts also contain 28.1% of all Fire Department facilities (16 out of 57).

At first glance, this seems reassuring. The areas experiencing the most harm also have the strongest emergency presence. But this is where the insight becomes counterintuitive: even though these districts have more police and fire coverage, they still face the highest levels of disorder and violence.

This creates a structural imbalance: the city concentrates responders where harm occurs, yet the upstream agencies that could reduce harm are largely missing. 

Why Prevention Matters

Preventive agencies shape the conditions that stabilize neighborhoods:

• Human Services and Homelessness Support reduce encampments and cycles of vulnerability.
• Public Health intervenes early in behavioral crises that might otherwise emerge as disorder.
• Public Works and the Public Utilities Commission (PUC) influence day-to-day cues through lighting, sanitation, and infrastructure.

These systems form the long-term safety net that allows neighborhoods to stay stable—not just respond to emergencies.

Yet the districts with the highest crime rates lack many of these foundational services.

Where the Gaps Are Most Severe

• Human Services Agency (HSA): No facilities currently operate in high-risk districts, leaving a 100% service gap.
• Public Utilities Commission (PUC): Population needs indicate about 21.3 facilities, but only 5 are in place.
• Homelessness & Supportive Housing: Capacity is also far below district need.
• Other agencies with notable deficits include Public Health, Public Works, and the Municipal Transportation Agency (MTA).

The Impact on Access

The chart below highlights how preventive infrastructure is overwhelmingly concentrated in lower-risk districts. For several agencies, high-risk districts hold only a small fraction of available facilities.

Acknowledging these gaps makes it obvious that the current system wasn’t designed for the complexity it faces. The path forward requires more than response, it requires alignment.

Conclusion: As per Gemini 3 and ChatGPT 5.1

We uploaded our research into both the cutting edge LLMs and asked them to weigh in. We asked the exact same question to both – “As an expert on urban planning and policy, weigh in with your closing arguments on this analysis.”

Here what Gemini responded:

San Francisco’s data reveals a critical structural flaw: the city has efficiently deployed its reactive forces to high-crime areas while leaving them devoid of preventive support. These "Convergence Zones" are overwhelmed not just by crime, but by a total collapse of administrative and social infrastructure, creating an environment where emergency responders are managing issues caused by the complete absence of agencies like Human Services. The path forward is clear: the city must stop treating these districts solely as crime scenes and start stabilizing them by physically embedding the missing preventive services directly where the strain is highest.

Here’s what ChatGPT had to say:

San Francisco’s data points to a simple truth: the neighborhoods carrying the city’s heaviest mix of crime, disorder, and service demand are also the ones with the weakest access to preventive systems, leaving responders to shoulder problems that upstream agencies were never positioned to solve. The patterns in this analysis make the path forward clear—treat these convergence zones as places where the city must rebuild its foundation, not just reinforce its perimeter. Only when long-term supports stand alongside emergency teams will these districts have a real chance to move from strain to stability. 

What do we have to say:

We aren’t policy experts and do not want to pretend like it as well. Our sole attempt was to answer our own curiosity and see a picture that perceptions might hide and analysis might reveal. Petavue is a truth-first analysis platform. Petavue is a truth-first culture.