San Francisco · March 2026 briefing

San Francisco Crime Rate Trends

The city, by the numbers we publish each month: 41 neighborhoods, 10 incident categories, twelve months of trailing comparison. Browse the rankings, scan the multi-year trends, or open a neighborhood-level breakdown.

Read methodologyUPDATED · AS OF 2026-03
207
tracked signals this month — across spikes, drops, sustained shifts, rare events.
-25.0%
overall city incident volume vs. trailing 12-month avg.
41
neighborhoods covered. Each gets its own page.
10
incident categories tracked, NIBRS-aligned.

Noe Valley other larceny is the sharpest single signal this month — a fresh spike standing out against a briefing otherwise dominated by below-trend readings. It is the lone spike in the top five; the remaining four entries are all drops, spread across vandalism, burglary, and motor vehicle theft in South of Market, Hayes Valley, Portola, and North Beach.

Citywide volume is down 25.0% against the prior 12 months — 37,408 incidents against 49,865 the year before. The signal mix reflects that structural decline: 100 sustained-shift signals and 91 below-trend readings across 41 neighborhoods, against a single spike in 207 total signals. South of Market vandalism and Portola burglary both ran below trend, consistent with the broad-based pattern.

This month's briefing is largely a continuation of what the prior several months established: a citywide decline held in place by widespread sustained shifts, with 15 zero-event signals and only one fresh above-trend move to track. The Noe Valley other-larceny spike is the one thread worth watching into April — everything else fits the existing arc.

WHAT TO READ NEXT
Direct answers

San Francisco Crime Frequently Asked Questions

Trailing 12 months vs the prior 12 months, computed from the same NIBRS-aligned categories used everywhere else on the page. Updated each March 2026 briefing.

Is crime in San Francisco down?

Yes — citywide incident volume is 25.0% lower than the prior 12 months.

Across the trailing 12-month window we tracked 37,408 incidents in NIBRS-aligned categories, compared to 49,865 in the year before — down 12,457 incidents.

Is violent crime in San Francisco down?

Yes — homicide, robbery, aggravated assault, and sexual assault are down 22.4% combined in the trailing 12 months.

That's 4,239 violent incidents in the past year against 5,461 in the prior year. See the by-category section below for the per-bucket breakdown.

Is property crime in San Francisco down?

Yes — burglary, theft from vehicle, larceny, motor vehicle theft, and arson are down 26.2% combined in the trailing 12 months.

That's 27,353 property incidents in the past year against 37,073 in the prior year.

Which neighborhood in San Francisco saw the biggest crime drop?

Japantown — 48.3% fewer incidents than the prior 12 months.

Japantown logged 187 incidents in the trailing 12 months against 362 the year before.

Which neighborhood in San Francisco saw the biggest crime increase?

Lakeshore — 10.3% more incidents than the prior 12 months.

Lakeshore logged 877 incidents in the trailing 12 months against 795 the year before.

City profile

The denominators behind the numbers

SOURCES · US CENSUS ACS 2023 · DATASF · SFPD
Geography
Land area46.9 mi²
Water area185 mi²
Coastline3 sides · Pacific + Bay
Elevation0–925 ft
Police districts10
Neighborhoods41 (analysis units)

Peninsula city; sharp microclimate and topography breaks (hills, parks, freeways) define neighborhood boundaries more than political ones.

Population
836,321
Density~17,832 / mi²
Median age39.7
Households~363K
Avg HH size2.24

ACS 2023 5-year estimates, county-level (San Francisco County).

Housing
Units~412K
Median rent$2,419
Median home value$1.38M
Vacancy11.9%
Tenure
Renter 62%Owner 38%
Stock
SFH 31%2–4 unit 20%5+ unit 49%
Economy & people
Median HH income$141,446
Poverty rate10.6%
Unemployment5.6%
Bachelor's+60.1%
Foreign-born34.2%
Age distribution
<18 14%18–34 28%35–64 42%65+ 17%

City-level only. We deliberately do not juxtapose these with neighborhood-level crime data — see the methodology for why.

Built environment
Street miles1,260
Parks (acres)5,860
Transit stops3,420
Walk score88 (high)

High pedestrian + transit exposure shifts crime-rate denominators away from “per resident” toward “per person-hour outdoors.”

Policing context
SFPD sworn officers1,537
Officers / 10K res.19.0
911 calls / yr1.21M
Open data lag≈ 48 hrs

SFPD reclassified some larceny coding in 2024; that window is noted on the methodology page so it doesn’t skew baselines.

Interactive map

Every neighborhood, color-coded

CLICK A NEIGHBORHOOD →
Category
Layer
Window
RAW COUNT · 1Y
231,8923,761
Rankings

Largest moves this month

SORTED BY |Z|
#NeighborhoodCategoryMoMYoY 12mo|z|90-day trendSignal
01Noe ValleyOther Larceny-6%+1%3.73SPIKE
02South of MarketVandalism-19%-19%5.75DROP
03PortolaBurglary-58%5.07DROP
04Hayes ValleyVandalism+22%-20%4.70DROP
05North BeachMotor Vehicle Theft-50%-44%4.62DROP
06Noe ValleyTheft from Vehicle-50%-49%4.58DROP
07West of Twin PeaksVandalism-43%-34%4.53DROP
08Nob HillVandalism-25%-24%4.49DROP
09Bernal HeightsTheft from Vehicle0%-53%4.35DROP
10Sunset/ParksideVandalism0%-14%4.32DROP
11Pacific HeightsVandalism0%-20%4.16DROP
12Bayview Hunters PointVandalism-3%-16%4.12DROP
Showing top 12 of 20 (neighborhood × category) cells with tracked signals.
Multi-year trends

The long arc — eight years of monthly counts

SELECT A CATEGORY ↓
036912201820192020202120222023202420252026monthly12-mo rolling mean
Latest 12mo57
YoY 12mo+27%
5-year change+338%
Window change+200%
Peak (12mo avg)5 · Jul '24
Trough (12mo avg)1 · Nov '20
ALL CATEGORIES · 8-YEAR ARC · 12-MO ROLLING MEAN
2018 ─────────────────── 2026
When does it happen?

Hour-of-day, day-of-week, and seasonality

Distribution of bucketed incidents citywide across the full analysis window. Useful for routine context — shopping-strip thefts vs. late-night assaults read very differently when you can see when each typically happens.

HOUR OF DAY · ALL CATEGORIES
020,00740,01512am6am12pm6pm11pm

Hour 0 is mildly inflated by reports without a known time defaulting to midnight — see methodology.

DAY OF WEEK · ALL CATEGORIES
046,03692,072MonTueWedThuFriSatSun
MONTH OF YEAR · ALL CATEGORIES
026,65453,307JanFebMarAprMayJunJulAugSepOctNovDec
Methodology

Every signal, every forecast, documented

Open about how we define spikes, what we exclude as noise, where the data comes from, and how often the model is wrong.

# anomaly rule — spike
flag = (z >= 2.5) AND (current_12mo >= 20) AND (current_6mo above sustained band)
where z = (current_12mo − μ_baseline) / σ_baseline
# exclusions (excerpt)
· simple assault (varies by reporting practice)
· drug offenses (reflect policing policy)
· admin records, weapons-possession, fraud
# 2025 backtest (citywide)
7 of 10 categories ≥ 90% coverage. see table →