Police should track mental health, homeless calls better, new audit says

The 80-page analysis — the first known assessment of its kind — was published Thursday by city auditor Jenny Wong.

Officials are looking to limit the nonviolent calls involving homelessness that police handle — part of the ongoing process to reimagine policing in Berkeley. Above: Officers respond to a dispute involving residents of neighboring downtown Berkeley campsites, July 22, 2020. Credit: Emilie Raguso

The Berkeley Police Department should do more to track calls involving mental health crises and homelessness and make more of its data available, according to a comprehensive city audit that has just been released.

Read more about policing in Berkeley on Berkeleyside

The 80-page analysis, which was published Thursday by city auditor Jenny Wong, provides the first detailed look perhaps ever undertaken into BPD calls for service, which include both reports from community members and activity initiated by police officers themselves.

“I’m hoping that the data will guide us to some successful pilot programs and help us make smarter policy decisions with a fairer use of resources,” Councilmember Ben Bartlett told Berkeleyside. It was Bartlett who called for the auditor’s analysis during budget talks last year related to how the city might reimagine policing and other city services in Berkeley in the wake of George Floyd’s murder and the widespread protests that followed.

To conduct the audit, Wong and her staff reviewed five years of Berkeley police data — about 360,000 “events” from 2015-19 — in an effort to better understand how officers spend their time and what duties might be shifted to civilian staff as part of the reimagining process.


Last year, city officials said any significant changes to the department budget should be backed up by reliable data.

Still no answer as to how many mental health calls there are

Police at the scene of a homeless fatality on San Pablo Avenue, October 2020. Credit: Citizen reporter

One of the biggest questions that has been raised in Berkeley over the years amid the myriad conversations around policing has been what portion of the agency’s calls for help involve people in mental health crisis or experiencing homelessness — or both. Unfortunately, the auditor’s office said it was not possible to answer this question accurately due to insufficient data.

This issue has become even more pressing in the past year as city officials and many community members alike have said they would like to reduce the police budget to increase resources for mental health workers and other social services. The city is looking at a pilot program, for example, to create a new Specialized Care Unit where clinicians and other health staff, rather than police, would respond to nonviolent calls involving mental illness.

In 2015, police estimated that at least 35% of BPD’s calls involve someone in crisis.

The auditor’s office assessed the new dataset using a rather complex narrative-based analysis — well explained in the report’s methodology — to identify these calls within the records. That was necessary because BPD does not have a consistent tracking system for these calls, according to the report.

The auditor’s office found that at least 12% of the calls it studied had a mental health component, according to the available data, and that at least 6% referenced homelessness in some way. (The categories overlap, the auditor noted.) But these numbers are unreliable.

“We identified as many of these events in the data as possible, but they are undercounted, likely significantly, because BPD does not identify all calls related to mental health or homelessness,” the auditor wrote.

As a result of this deficiency, the primary recommendation in the report is for BPD to identify all of these calls — while respecting privacy — in a way that can be assessed and tracked.

“City Management agreed to our findings, conclusions, and recommendations,” according to the report.

In a brief response to the audit that covers less than two pages, and was described by the auditor as an “initial corrective action plan,” the Berkeley Police Department said this change should be feasible but will require more training and better data tools. BPD said this change could likely be accomplished within six months.

The second recommendation in the audit is for Berkeley police to make more department data available online to the public for review and analysis. On the city’s open data portal, BPD currently posts calls for service covering the past six months; 30 days of arrest and jail booking data; and stop data going back to 2015.

Last October, BPD began posting much more detailed stop data on that website well in advance of the state’s 2022 deadline to do so. The department also committed to reporting on that data publicly, which has not happened yet.

In the new report, the auditor’s office asked BPD to broaden the categories it shares about all of its calls for service and to make years of that data, rather than months, available online “to facilitate transparency.” BPD agreed to do this, according to the report, but said it may take six months to a year to accomplish, depending on whether it needs to secure a new vendor to make the adjustments.

The auditor’s office also said BPD should post the full dataset underlying its new analysis on the city’s open data portal for review.

BPD did not respond this week to a request for comment about the auditor’s report, but it generally does not grant interviews about reports that will come to the City Council until after the public meeting takes place.

Racial disparities continue, Berkeley police audit shows

In addition to its broad look at calls for service, the auditor’s office also zeroed in on officer-initiated stops of motorists, cyclists and pedestrians. This portion of the stop data — records related to about 56,000 individuals stopped over five years — confirmed previously identified trends showing racial disparities in vehicle stops, with Black and Hispanic drivers searched more often than white drivers but less likely to be arrested after a search.

Vehicle stops were the most common stop type of this subset and most took place between 9 p.m. and midnight, the auditor found.

The most common outcome after an officer-initiated stop was a warning, about 60% on average across all races. The arrest rate across all races was also similar, with an average of about 3%. That rate was slightly higher for Black individuals (5%).

Black individuals who were stopped were about half as likely to receive a ticket than white or Asian individuals. This datapoint has been the subject of much debate.

Community members who have previously and vociferously expressed concern about these trends say the lower citation rate means police are stopping Black individuals for no reason.

Officers in Berkeley have said repeatedly in various forums over the years that this lack of enforcement action is actually evidence of how they use their discretion when they can to enact less severe penalties when stopping members of communities that have historically been subject to overpolicing.

In the past, police did not track data related to whether evidence or contraband had been collected as a result of these searches, which made it difficult — everyone agreed — to interpret the disparate outcomes. Berkeley police collect that data now under the new state rules.

The auditor’s office also called out the fact— as have many others over the years — that Black individuals are highly overrepresented in the stop data when compared to Berkeley demographics. These individuals make up just 8% of the city population but account for 34% of the stops. Meanwhile, white people make up 53% of the population but only 35% of the stops.

(The auditor’s office noted in this section of the report that multiple stops — one for each person detained — could be tied to a single call for service or event, and did not know as of publication time how many actual events these stops reflected. Update, April 29: The auditor’s office said the stop data related to about 53,300 events.)

Researchers have cautioned against using population numbers to draw conclusions about disparities, calling them the “crudest possible benchmarks,” because city borders are fluid and demographics can — and do — vary widely in nearby jurisdictions. It was not addressed in the audit, but many of the people who are stopped by police in Berkeley are not city residents. And, according to a recent BPD analysis, about 40% of the people who commit crime in Berkeley come from outside the city.

Despite these caveats, the population comparison figures continue to be highly cited.

City officials have already taken steps to set new policies in an effort to address the disparities, coming up with a plan that may one day remove armed police from traffic stops and replace them with unarmed civilians — assuming state law can be changed to allow this — and also reducing the type of stops police are allowed to make. Other efforts to reimagine policing are also underway.

Traffic stops, disturbances, alarm calls are most common

The most serious police incidents, such as this homicide investigation in October 2020, make up a tiny portion of the agency’s calls for service. Credit: Emilie Raguso

According to the auditor’s analysis, traffic calls for service are the most common call type documented by the department, with nearly 45,000 listed over the five years studied. The auditor’s office also elected to keep all traffic stops in the traffic category even if they resulted in a crime reported.

Disturbances were the next most common call type, with about 36,000 tallied over the five years studied, followed by alarm calls (20,000), noise disturbances (16,000), security and welfare checks (about 15,000 each) and parking violations (14,000). Suspicious circumstances, trespassing and theft calls each resulted in about 10,000-12,000 events.

In total, the top 10 call types made up 54% of all calls for service over the period studied. BPD averages about 72,000 calls for service each year.

Calls for service categorized by the auditor as felony violent crime reports made up less than 1% of the events in the dataset. The call type related to possible gunfire was not counted in that category, however, nor were investigations that may have related to serious crimes. But combining all of those statistics, as categorized by the auditor’s office, still results in just a fraction of the overall data.

Another 7% of the calls were classified as felony property crime reports with another 22% reflecting reports of less serious crimes.

It’s important to note that all crime reports that began as traffic incidents were excluded from the crime categories — they show up in the traffic bucket — making it difficult to understand how common crime calls in Berkeley actually are due to the way the dataset was analyzed.

One huge drawback with the dataset itself, however, is due to the nature of calls for service. Calls for service are only as accurate as the report being made; the information is completely unconfirmed. Take for example the unprovoked fatal shooting last year of UC Berkeley student Seth Smith. That incident was reported to BPD as a person down in the street.

Under the auditor’s classification, that would have shown up in the medical or mental health bucket rather than violent crime.

This inherent problem with the inaccuracy of calls for service is one reason law enforcement agencies do not spend a lot of time analyzing these incidents. The dataset is made up of the completely preliminary calls people make to police, and issues officers witness, but does not represent the verified crime or type of circumstance determined to have occurred.

The numbers that are more readily available from BPD and other agencies each year are a much smaller dataset that are collected as part of the FBI’s Uniform Crime Reporting (UCR) program. Those incidents are reviewed and vetted before they are assigned to eight categories such as homicide, robbery, burglary and rape.

These are the crimes the federal government tracks annually for law enforcement agencies across the nation and that BPD reports to the Berkeley City Council each year.

That UCR data has its own limitations, however, and includes no context about who may have been stopped, how a call originally came about or how long an officer spent on it. That information can only be found in the type of calls for service datasets the auditor’s office analyzed.

Next steps for audit of Berkeley police data

The new audit reflects work that began in response to George Floyd’s murder last year, officials said. Above: Black Lives Matter protesters hold a vigil for Floyd on June 13, 2020. Credit: Pete Rosos

Wong told Berkeleyside she filled an open position in her department, and also assigned her senior staff members to the audit, so there would be enough resources to tackle the complex analysis in a timely way that could aid in the important work the city is doing around police reform.

“I had my whole team working on this,” Wong said. “We asked the police department a lot of questions.”

Wong said it had been a challenge for staffers to ensure they understood the data and to analyze all 360,000 events in a way that would make sense. The resulting report also walks through all sorts of common law enforcement terminology and a range of information about Berkeley police resources to give readers the lay of the land.

Learn more about the auditor’s office on the city website

“Transparency is really critical to good government and good policing,” she said. “Ultimately this report is the first ever, that I know of, that looks at the review of calls for service and officer-initiated stops in Berkeley at a time when council, the department, the community — everyone — is looking at how to reimagine public safety in the city.”

The new audit will be presented to the city’s Reimagining Public Safety Task Force next week, on April 29, and will then go to the Berkeley City Council on May 11.

Councilmember Bartlett told Berkeleyside he had called for the analysis last year to ensure he and his colleagues on the City Council would be able to make data-driven decisions about the Berkeley police budget that would have a lasting impact. Officials have said they hope to focus BPD resources on violent crime and away from mental health calls involving people in crisis.

“In order to make that case, we needed this data,” he said. “We knew, if we did it in a knee-jerk fashion, the reforms would be reversed after we were gone.”

Bartlett said the auditor’s work had been “impressive,” and that he had been surprised to learn how limited BPD tracking is around mental health calls. He said officials will look to create new policies to close that gap.

Bartlett also said he had been surprised to see how few police stops result in arrests and that it seemed to provide justification for shifting many traffic stops to civilian staff that would work within “BerkDOT,” a new Department of Transportation proposal that is still in the concept stage.

He said it was also fitting that the Berkeley police audit was released this week just days after a guilty verdict in the George Floyd murder case against former Minneapolis police officer Derek Chauvin.

“It’s been a year, almost to the week,” Bartlett said. “George Floyd’s death caused us to have the big march that buried racism. We all came forward with our initiatives. It gave us the impetus to do the reimagining. To have his murder adjudicated right now, at this point where we’re actually landing our efforts that were born from his death, it feels like we are tied to his death — and that our efforts will be part of his legacy.”

Emilie Raguso is Berkeleyside’s senior editor of news. Email: emilie@berkeleyside.org. Twitter: emraguso. Phone: 510-459-8325.