I mapped 40 years of Santa Cruz crime data: here’s what it can’t tell you

CruzLens started as a crime map of Santa Cruz County. The goal was simple: put every public crime report on a map, going back as far as the records go. What we learned building it is mostly about what the records leave out.

There are two very different kinds of crime data here. Official annual totals go back 41 years, but say nothing about where. Incident reports say where, but only a fraction of them are public, for windows ranging from a month to a few decades.

Source: California DOJ OpenJustice, Crimes and Clearances with Arson 1985–2025 · Data: doj-santa-cruz.json · Code: CruzLens

Show the data (41 rows)

Year	Part I	Violent	Property	Arson	Agencies reporting
1985	15,611	1,085	14,450	76	11
1986	15,121	1,248	13,754	119	11
1987	14,324	1,146	13,073	105	11
1988	13,508	1,067	12,376	65	11
1989	13,486	1,023	12,413	50	11
1990	13,794	1,099	12,620	75	11
1991	13,775	1,253	12,454	68	11
1992	14,078	1,594	12,409	75	11
1993	14,138	1,704	12,344	90	10
1994	15,006	1,677	13,239	90	10
1995	13,940	1,757	12,086	97	10
1996	13,399	1,774	11,550	75	10
1997	12,285	1,613	10,577	95	10
1998	10,051	1,394	8,591	66	9
1999	8,761	1,065	7,612	84	9
2000	8,735	1,187	7,504	44	9
2001	10,171	1,203	8,910	58	9
2002	10,208	1,099	9,037	72	9
2003	11,187	1,268	9,865	54	9
2004	10,654	1,198	9,397	59	9
2005	11,459	1,155	10,255	49	9
2006	11,104	1,083	9,936	85	9
2007	10,194	1,239	8,866	89	9
2008	9,201	1,247	7,882	72	9
2009	10,341	1,215	9,059	67	9
2010	9,642	1,313	8,267	62	9
2011	9,952	1,169	8,744	39	9
2012	9,793	1,041	8,709	43	9
2013	9,202	1,002	8,140	60	9
2014	9,312	1,139	8,121	52	9
2015	10,502	1,171	9,282	49	9
2016	10,624	1,101	9,485	38	9
2017	10,072	1,056	8,947	69	9
2018	8,603	989	7,530	84	9
2019	8,165	1,106	7,005	54	9
2020	7,200	970	6,155	75	9
2021	6,604	1,052	5,478	74	9
2022	6,510	1,200	5,239	71	9
2023	5,018	1,070	3,886	62	10
2024	5,045	939	4,027	79	9
2025	4,644	862	3,728	54	9

The 41-year view: totals without places

The California Department of Justice publishes annual counts for every police agency in the state. For Santa Cruz County that is 391 agency-year rows from 41-year series like the Santa Cruz Police Department to short-lived ones like the Pajaro Coast state parks unit.

The long view is clear: reported Part I crime was 15,611 in 1985, the highest year in the series, and 4,644 in 2025. Almost all of that decline is property crime (−74.2%). Violent crime moved much less (−20.6%), with a rise through the mid-1990s:

Source: California DOJ OpenJustice, Crimes and Clearances with Arson 1985–2025 · Data: doj-santa-cruz.json · Code: CruzLens

Three things in the DOJ file make the trend harder to read than the chart suggests:

• Definitions changed. The rape definition broadened in 2014, and from 2021 the DOJ combines old summary-format submissions with converted incident-based (NIBRS) ones. A step at either year can be a counting change.
• Agencies come and go. 11 agencies reported in 1985 and 9 in 2025; Cabrillo College’s police stopped appearing after 1997.
• No denominators. The file has counts, not rates. The county’s population and its summer visitors changed over 41 years, and the totals don’t adjust for either.

And none of the totals have a location below the agency. A Sheriff’s count covers everything from Boulder Creek to La Selva Beach.

The incident view: places without completeness

To map crime you need incident records with addresses, and in Santa Cruz County those come from several sources, each with its own gap:

• Santa Cruz Police publishes approved activity logs as PDFs. Each covers about a day and is kept online for roughly 30 days, so a map built today only reaches back about a month. Our first snapshot held 485 reports from 32 documents.
• The Sheriff publishes case-log PDFs. The index page labels some of them with the wrong month (a July file labeled June), so dates are read from inside the documents. Six PDFs added 2,119 reports covering April to July 2026.
• Watsonville publishes scanned logs that need OCR. Of 20 indexed documents, 15 were readable; five failed our validation checks and are shown as gaps rather than guessed at.
• UC Santa Cruz and Scotts Valley have public case maps going back to 1997 and 2012, which added 5,399 and 719 cases.
• Capitola’s public map refuses queries older than 12 months; it added 65 reports.

Put together, the sample is dominated by whoever publishes the most:

Source: SCPD, Sheriff and Watsonville logs; UCSC, Scotts Valley and Capitola public maps · Data: cruzlens.json · Code: CruzLens

Show the data (12 rows)

Community	Mapped reports
Santa Cruz	6,196
Scotts Valley	749
UC Santa Cruz	456
Watsonville	408
Aptos	252
Soquel	196
Unspecified	145
Felton	127
Boulder Creek	120
Capitola	80
Ben Lomond	61
La Selva Beach	20

The city of Santa Cruz accounts for 70% of the 8,810 mapped records, and Watsonville for 408. That is not because Watsonville has little crime: over 2016–2025 its police reported 13,471 Part I crimes to the DOJ. It is because its logs are scanned, short-lived and partly unreadable. A gap on the map means missing records, not safety, and CruzLens draws unknown coverage in gray instead of green.

Wrangling PDFs without leaking people

Police logs are written for people, not machines. Turning them into rows meant:

• reading text positions from the PDF rather than eyeballing columns, and falling back to OCR for scans;
• keeping only the location and charge columns, and dropping names, ages and narrative text at parse time;
• deduplicating cases that appear in several daily documents, keeping the newest version;
• placing each report at the center of its block, never at an exact address, and leaving unmatched locations unmapped rather than guessing.

That parser became its own tool, blotter2csv, which drops personal names and narratives by default.

What the data can’t tell you

• Whether a neighborhood is safe. Reports reflect what was reported and published, not what happened, and publication windows differ by agency from a month to decades.
• Trends at the block level. The long series are countywide agency totals; the block-level records mostly cover the last year.
• Comparisons between communities. Different agencies, sources and windows make rates meaningless across places.

What it can do is show what the public record contains, link every report to its source and make the gaps visible. That is why the current CruzLens default view is a map of crashes, spills, water quality and other conditions of places, with crime as one small, optional layer. The design rules behind that map are in a separate post.

Methods

• Annual totals. California DOJ "Crimes and Clearances with Arson, 1985–2025", all Santa Cruz County agencies summed by year (scripts/derive_doj.py). Part I = violent (homicide, rape, robbery, aggravated assault) + property (burglary, vehicle theft, larceny-theft) + arson. Negative or blank published values are treated as unknown, not zero. Our 2016–2025 total for the Santa Cruz Police Department (28,606) matches CruzLens exactly.
• Incident sample. Counts by published community label from the CruzLens police snapshot. Source descriptions, document counts and OCR failures are from the CruzLens README and docs/SOURCES.md.

Limits

The DOJ uses a hierarchy rule (generally only the most serious offense per incident is counted, with arson separate), so totals undercount offenses. Agency boundaries do not match community names. The incident sample is a snapshot and changes every time logs expire or new ones are published.