Last week the Seattle Police Department published its annual report on stops and detentions, with numbers for 2018. It begins by noting that Terry stops are up substantially over 2017 numbers — but then it provides another 25 pages of text, tables and charts that offer little help in understanding why. In fact, SPD itself admits in the report that they don’t know what the increase means:
SPD conducted 18.5% more Terry stops in 2018 than in 2017. However, statistical trend analysis shows that this increase appears to be an anomaly. Our analysis next year will indicate whether a pattern is developing or whether 2018’s data are a one-time increase.
But the data tells a clearer story than SPD is letting on. It took a few days of work to reverse-engineer and otherwise piece together the data that the department left out of its report and to follow the bread crumbs through it, but here’s the story that the data tells.
First, a refresher: a “Terry stop” is when a police officer temporarily detains someone for questioning, and possible frisk/search, based upon reasonable suspicion that the person had been, is, or soon will be involved in criminal activity. Courts have said that such stops (and searches) are legal so long as the officer can articulate the source of the reasonable suspicion. However, the broad use of Terry stops is often associated with biased policing when it can be shown that officers stop certain demographic groups (e.g. people of color) more often than others — which can be interpreted by the people stopped as a form of police harassment. SPD collects data and reports annually on its use of Terry stops as one of the conditions of the consent decree with the Department of Justice. It also reports out on the number of times an officer frisks a suspect during a Terry stop, the number of times the frisk results in finding a weapon, and the number of times the Terry stop results in an arrest.
Terry stops are not the same as “voluntary” stops, where the person that an officer is interacting with is free to leave at any time. During a Terry stop, the person is in the custody of the officer until released or arrested. Terry stops are also not the same as arrests: an officer does not have “probable cause” to arrest and charge the person with violating the law, but can hold them in custody long enough to ask them questions and frisk them — either of which might lead to probable cause and an arrest.
There is no “right” number of Terry stops that a police department should be doing. Rather, we can think about how the volume of Terry stops relates to other ways that we measure the performance of a police department, such as:
- Are the Terry stops an effective tool? If the number of Terry stops go up, then do the number of arrests go up as well? If officers choose to frisk more people that they have stopped, do they find more weapons, narcotics or other contraband?
- Are increases in Terry stops tied to reasonable principles of when and where they are expected to be effective, or are they used haphazardly or targeted at specific demographics without evidence that those targets deserve more scrutiny?
- What are the negative impacts of Terry stops upon individuals and groups who are targeted by them, and how do they affect the overall perception of and trust in the police?
Here is SPD’s top-level numbers of Terry stops in 2018:
All told, Terry stops increased by 18.5% from 2017, with the increase concentrated in three precincts: East, West and South. East Precinct had by far the biggest increase, both in absolute numbers and by percentage.
According to the report, SPD is at a loss for what caused this big spike that happened unevenly across the city, calling it an “anomaly.” And yet, a chart of the number of stops (and frisks, and arrests) per month shows something interesting:
In May 2018, there was suddenly a spike in Terry stops, with corresponding increases in frisks and arrests. What happened last May that would cause this? The department provides no answers, but a few hints. One clue is its observation that an increasing supermajority of Terry stops happened on dispatch calls, versus officers’ own initiative:
That suggests we should dig into the data on calls to SPD to look for something that happened around last May. Overall, the month-to-month call data looks fairly flat, with a small tick up in May:
However, that increase isn’t spread evenly across types of calls. I found three with particularly large jumps around May: weapons, assaults, and “shots fired.”
By volume, the assault calls are the big deal, but the clearest spike is for shots fired.
Interestingly, of the three, only “assaults” makes SPD’s list of call types that directly led to high numbers or bigger increases of Terry stops. That suggests that if these calls were the trigger, they led SPD to implement a broad increase in stops, and not just for these particular kinds of calls.
We can dig in further and look at the precinct-by-precinct distribution of the calls to see how they match up with the increases in Terry stops, which you will recall were concentrated in the East, West and South precincts. You can see that there were May increases in assault calls in the East and South precincts, as well as the Southwest:
Increases in shots fired calls in May, however, were concentrated in the North and Southwest:
It’s well documented that crime incidents increase during the summer months (nationally, not just in Seattle). This data suggests that SPD was seeing an upswing in calls related to assault, shots fired, and weapons-related crimes starting in May, and responded in kind with increased Terry stops — though its focus areas didn’t clearly match the precincts where the calls were coming from. This may mean that the department’s response was uneven or poorly targeted. However, it may instead mean that SPD officers are aware of well documented patterns of under-reporting and over-reporting certain crimes, and they take that into consideration when responding to calls in neighborhoods with a history of under-reporting or over-reporting. This is one of many cases where the data doesn’t tell us the “why” behind the observable changes, and more investigation is required.
We can look further into the data to see signs of other disparities. For example, we can look at the ages of the people stopped. Across the whole city, the distribution looks like this:
You can see that the increase in Terry stops is heavily concentrated in the 26-45 age range. But again, that isn’t uniform across the city. South and West precincts generally match the city as a whole, and North precinct as we know didn’t see any growth:
But the Southwest precinct, despite having no total growth, skewed its Terry stops much older:
And the East precinct showed increased Terry stops in all age ranges. In fact, it is the only precinct that saw increases in Terry stops for people age 25 and younger.
It’s difficult to know what these changes mean, though we do have some context. Recently we found out that SPD has been conducting “emphasis patrols” through its gang unit in East Precinct for some time; that aligns with the data showing increased stops of people who would be in the right age to be potential gang members. However, SPD hasn’t provided data on gang unit activity, so we can’t understand how that maps onto the larger Terry stop trends.
We can also look at racial disparities. SPD breaks out the race-related data separately for women and men (I scaled the two charts the same to make it easier to compare them):
Overall, it skews heavily male, and is almost entirely two racial groups: blacks/African Americans and whites. Blacks continue to be over-represented in the Terry stops data compared to their representation in the entire population of Seattle. And we see the same demographic distribution even stronger in the East Precinct:
At this point, we don’t have enough information to know whether these disparities are warranted; we need to see the results of the stops, and to do that we need to examine data for frisks, weapons found, and arrests.
When we look at the data for frisks, we see that the focus is really on just two precincts: the East and South. The number of frisks during Terry stops increased in only those two precincts, and the likelihood that a stop would lead to a frisk also increased in those same two places:
Interestingly, though perhaps not surprisingly, the increased frisks did indeed lead to more weapons being found in East and South precincts
It increased especially in the East Precinct, and the rate of finding a weapon during a frisk (what SPD calls the “hit rate” for a frisk) also substantially increased there. The hit rate also went up in the West Precinct. It’s of note, however, that the hit rate for a frisk went down in the North and South precincts. That’s particularly eyebrow-raising for the South precinct, where the number of Terry stops and frisks increased substantially, but their effectiveness decreased. On the flip side, it looks like a big success for the East Precinct: Terry stops and frisks increased, a lot more weapons were found, and the hit rate for frisks more than doubled.
And also not surprisingly, that led to more arrests from Terry stops in the East and West Precincts, and higher rates of arrests from stops in those districts as well. However, SPD’s report does not provide any data on age or racial demographics for arrests, so we don’t know if there are any disparate impacts (though given the data on stops, there almost certainly are).
Despite SPD’s claims that the 2018 numbers are an “anomaly,” there is ample evidence of intentional action, potentially driven by increasing calls for service related to shots fired, assaults and weapons use — though to be clear this is correlation, not proven causation. SPD’s response had its greatest impact on the East Precinct, with scattered impacts in the other four precincts. There are continued disparate impacts on black residents, particularly in the East Precinct, and uneven use of Terry stops on people of different ages across the city with a heavy focus on younger males in the East Precinct.
That said (and putting the disparate impacts aside for the moment), the increase in total stops and frisks alone does not necessarily mean that SPD is doing something wrong — and in some parts of the city it appears to be a sign that they are doing something right. They found more weapons, both in absolute numbers and as a higher percentage of the frisks performed. They made more arrests. And particularly in East Precinct, the increased Terry stops and frisks are showing clear results. But more broadly across the city, without longer-term data and any sense of what a “normal” or “expected” rate of effective Terry stops should be, two years of data doesn’t tell us whether this represents a good or bad policing practice.
With regard to the stops and detentions report and its claim that 2018 was an “anomaly,” it’s hard to tell whether SPD truly is unaware that this is going on, or is aware of it but chose to obfuscate it at a particularly sensitive time for the Consent Decree. The 2017 stops and detentions report provided deep statistical analysis of stops, arrests and seizures, and explicitly highlighted many of its findings. But last year it abandoned that approach and produced a slimmed-down version with much less useful analysis and attempts to glean meaning from the statistics. This year the department invented a whole new set of charts and graphs for its report that it didn’t use in either of the previous years, and yet most of them either miss the important data or present it in a manner that hides anything insightful. To illustrate the point, here are some of the most infuriating charts from their report:
Here they chose to use the height of the bars to represent the percentage (out of 100%) of the total subjects in that age range. If you look closely at the 18-25 age range, you’ll notice that the 2018 number is greater than the 2017 number — but the bar is shorter. Compare that to my re-rendering of the chart:
Likewise for the East Precinct, which completely obscures the dramatic changes in absolute numbers. This might be excusable if they actually charted the absolute numbers too, but they didn’t — this was all we got.
Here’s my version:
And then there’s this beauty, where the bars represent the percent change from the previous year — completely obfuscating the relative size of the changes.
My redos:
And finally this graph for the racial distribution of Terry stops in the East Precinct, where the large percentage increase in a very small number obscures everything else. The absolute numbers given represent the increase, but they don’t even bother to tell us what the baseline is.
My version, which tells a completely different story with the same data:
Again, it’s difficult to tell whether this is intentional obfuscation on SPD’s part. Either way, it’s very effective at throwing people off the scent. There is evidence that SPD knows how to do this kind of analysis effectively: in January its audit department produced a report concluding that SPD officers may be over-reporting Terry stops — though since those procedures have been in place for some time, that doesn’t explain why there would be a sudden increase in Terry stops, frisks and arrests beginning last May. SPD also published the first part of its disparity review in April, which does a deep-dive on known disparities (including for use of force) and applies a fairly new concept called “Propensity Score Matching” to analyze whether those disparities may be warranted. Clearly SPD has some muscles in this area; why they chose not to use them in the case of its most recent stops and detentions report is baffling.
There are still some important questions to be answered before we can fully understand the most recent data on Terry stops, frisks, and arrests. Among them:
- Did SPD explicitly direct officers to start making more Terry stops last May, in either part or all of the city? If not, why did officers across the department suddenly increase their stops, frisks and arrests? Was that response proportional to what was happening at the time?
- When did SPD’s gang unit “emphasis patrols” begin, and what was their contribution to the increase of Terry stops, frisks and arrests — especially in East, South and West precinct?
- Is the increased number of Terry stops of people under age 25 in the East Precinct tied to the increased frisks, weapons found, and/or arrests in that precinct?
- Are there age and racial disparities for the frisks and arrests, and do they differ from the known disparities for Terry stops?
- When SPD officers conduct Terry stops, what are the reasonable suspicions they assert?
- What were the results of the arrests that arose from Terry stops, in terms of criminal justice outcomes?
- Was the increase in Terry stops an effective response to crime in Seattle?
- Has the increase in Terry stops influenced the public’s perception of and trust in SPD?
Fortunately, thanks to the police accountability legislation passed in 2017, we now have a Community Police Commission and an Office of the Inspector General that can dig into these questions.
SPD has made a commendable public commitment to transparency, including making much of the data they collect public. They publish several data dashboards and data sets, including the Terry stop data and a dashboard on calls for service to SPD that was invaluable in researching for this article. However, they don’t always live up to the promise: I downloaded the raw Terry stop data set they provided into Excel, but I discovered that the data is both incomplete (the totals don’t match SPD’s published report) and “dirty” in that fields are missing in large numbers of entries; in the end, it was of little use to me in validating and extending SPD’s report. Also, several of the dashboards can be difficult to navigate, making it hard to find the information you’re looking for. The department should be applauded for its public commitment to transparency and the steps it has taken so far, but it still has work to do to reach its transparency goals. Hopefully its recently-installed Data Access Platform and recordkeeping system will improve the quality of their data and help the department to share it more easily.