“The second time I watched it with the sound off.” A pause. “You know what I kept thinking? You’ve got the exact same thing I do, where you can’t back out of a room once you’ve walked into it. Your grandfather had it. I’ve got it. I put a hole in a drywall at the union hall in 1994 over a parking space.” Another pause. “Nobody ever told me I could just leave a room, Ry. Nobody ever told you either. I’m sorry about that.”
Caldwell did not trust his voice, so he said nothing.
“Anyway,” his father said. “I’m not going anywhere. Call your mother back, she’s pretending she’s not upset.”
On day nine, Caldwell received a letter.
Not an email. A letter, in a hand-addressed envelope, forwarded from the precinct.
It was one page, typed, unsigned except for initials, and it read:
Officer Caldwell —
You don’t know me. I was at the park. I’m the one who was asleep on the bench with a book on my chest, about forty feet behind you.
I woke up when you started yelling and I watched the whole thing and I did not do anything. I didn’t film it. I didn’t say anything. I just lay there with my eyes open pretending to still be asleep because I didn’t want to get involved, and when the trucks came I got up and walked to my car and drove home.
I’ve thought about that for nine days. I’m fifty-one years old and I lay on a bench and pretended to be asleep.
I’m not writing to make you feel better. What you did was wrong and I think you know that. I’m writing because I’ve read about four hundred comments now from people who are absolutely certain what they would have done in that park, and I was in that park, and what I did was nothing.
I don’t think most people are as sure as they sound.
— R.M.
Caldwell read it eleven times.
He put it in the drawer of his nightstand, and he kept it, and four years later it was still there. ## Chapter Seven — What the Files Said
The investigation took eleven months.
Almost none of it looked like anything. There were no raids. There were no dramatic subpoenas served on camera. There was a rented floor of a downtown office building with bad carpet, sixteen people, four hundred and nine boxes of paper, and a server.
Dr. Chidinma Okonkwo ran the data side. She was forty-four, a statistician by training, and she had spent nine years in academia before concluding that she would rather have her findings change something than be cited.
She described the work, later, as “archaeology.”
“People imagine we went looking for a smoking gun,” she told an interviewer. “There is no gun. There is never a gun. What there is, is four years of stop data, and when you line it up and ask it the right questions, it tells you what the department actually values — not what it says it values, not what’s in the policy manual. What it does ten thousand times a week without thinking about it.”
Her team asked the records eleven questions.
The first was simple: who gets stopped?
The answer, controlling for the residential population of each precinct, for the daytime population, for crime reporting rates, for calls for service, and for four other variables that took three months to construct properly, was that a Black resident of the city was 3.4 times as likely to be stopped on foot as a white resident, and 2.1 times as likely to have a vehicle searched after a stop.
The second question was the one that mattered more, and it was the one almost nobody outside the field thinks to ask.
What did the searches find?
Searches of white drivers produced contraband 21 percent of the time.
Searches of Black drivers produced contraband 11 percent of the time.
Searches of Latino drivers produced contraband 9 percent of the time.
Okonkwo put that on a single slide with no commentary when she briefed the city council, and then she stood next to it in silence for a count of fifteen, which she had practiced in a hotel room the night before.
Then she said:
“I want to be precise about what this means, because it is often described badly. This does not mean officers are searching Black drivers who are more innocent. Every one of these people is presumed innocent and most of them are. What it means is that officers required less evidence before deciding to search a Black driver. They searched on a weaker hunch. The hit rate is a measurement of the threshold. The threshold is not the same for everybody, and the threshold is the thing the Fourth Amendment is actually about.”
A councilman asked whether that could be explained by deployment patterns — more officers in higher-crime areas, therefore more searches.
“Yes,” Okonkwo said. “That is a real effect and it is the correct question. We modeled it. It accounts for about a third of the gap in stop rates.” She clicked to the next slide. “It accounts for none of the gap in hit rates, because the hit rate is conditional on a search already having happened. The denominator controls for deployment automatically. That’s why we use it.”
The councilman, to his credit, said, “Okay. Thank you. That’s a good answer.”
The second strand of the investigation was the narratives.
Every stop generates a few sentences of free text. Over four years, that is roughly one point one million sentences.
Three analysts read a stratified random sample of eleven thousand of them. It took five months. Two of the three said afterward that it was the worst work they had ever done, not because of brutality — there was very little brutality in the text — but because of the sheer volume of nothing.
They built a taxonomy of justifications. The top five, by frequency:
- Subject appeared nervous upon seeing marked unit. — 18 percent
- Subject was present in a high crime area. — 16 percent
- Subject made furtive movement. — 14 percent
- Subject appeared to be checking parked vehicles. — 9 percent
- Subject matched general description of suspect from prior call. — 7 percent
Caldwell’s own cards used phrase number one 94 times and phrase number four 31 times.
Arthur Bellamy was phrase number four.
The report’s language on this point became, briefly, a minor sensation among lawyers:
These are not observations. They are categories. An observation is specific to a person; a category is a container into which a person is placed. “Subject appeared nervous upon seeing marked unit” describes, if it describes anything, a universal human response to the sudden presence of armed authority. It cannot simultaneously be the universal baseline and the indicator of deviation from the baseline. Yet in this Department it functions as both, and functions so routinely that it appears, in some form, in nearly one in five documented stops.
An articulable suspicion that applies to everyone articulates nothing.
The third strand was the complaint system.