In the late 1990s, Los Angeles Police Department detectives in California uncovered one of the biggest policing scandals in modern U.S. history. More than 70 officers in the antigang unit of the city’s Rampart Division were accused of stealing drugs, robbing banks, beating suspects, framing defendants, and—in a separate civil suit—conspiring to murder rapper Christopher Wallace, aka The Notorious B.I.G. Researchers have long studied patterns of misconduct in such “rogue units,” to see whether they can identify how misbehavior spreads from one officer to another. Now, a team of economists says it has gone even further and measured, for the first time, the influence that misbehaving officers have on one other.
The approach is “promising,” says Robert Worden, a political scientist who studies criminal justice at the State University of New York in Albany, who was not involved with the work. But he’s skeptical that it reveals anything about how police misconduct really spreads. “It’s really hard to pull together that kind of data,” he says. “There’s no magic bullet.”
To find out how co-workers and friends influence each other, researchers typically look for patterns in the data: If a person has many obese friends, for example, there’s a good likelihood that they will gain weight. Dozens of studies have uncovered this kind of “peer network” effect for everything from smoking rates to test scores to productivity. (One study shows a whopping 12% increase in productivity when workers are surrounded by hard-toiling colleagues.) But these studies all share a weakness—even though they can show how people change in the presence of others, they cannot ultimately explain what caused the changes.
Edika Quispe-Torreblanca didn’t set out to pin down that causal link. Instead, the economist at the University of Oxford in the United Kingdom—then a graduate student at the University of Warwick in the United Kingdom—was invited to look for trends in 4 years’ worth of data from London’s Metropolitan Police Service. When she and her adviser, University of Warwick economist Neil Stewart, asked the deputy commissioner what the department worried about most, the answer was simple: misconduct.
“Legitimacy is crucial for the police,” Quispe-Torreblanca says. Stewart puts it more bluntly: Everybody—janitors, electricians, professors—cheats a little bit. But, “There aren’t many people with the power to imprison somebody. If you’re going to give people that kind of power, you don’t want them abusing it.”
To find out how “bad apple” officers affected their colleagues’ behavior, Quispe-Torreblanca and Stewart pored through the personnel files of 35,924 officers and staff, of whom nearly 15,000 had at least one complaint lodged against them. Some complaints were serious. Others involved improper logging of evidence or use of computer databases to do things like “looking up your sister’s new boyfriend to see if he’s a decent sort of chap,” Stewart says.
To show true causation, the pair couldn’t simply log all cases of coincident bad behavior. It’s possible, for example, that staffers with certain personality traits would gravitate toward certain units, or that a specific environment—such as a high-crime neighborhood—might make officers more likely to engage in misconduct.
So the researchers came up with what Stewart calls an econometric “trick.” For every person in the database who had served with different sets of colleagues, the researchers looked at two things: the numbers of complaints lodged against the person’s first and second group of co-workers. If the numbers were higher than the average across the entire data set and went up in lockstep as an officer moved from the first to the second group, the researchers could assume the relationship was causal—and that the linking officer was somehow responsible.
Sure enough, the scientists found that for every 10% increase in complaints against the first set of colleagues, complaints against the second set of colleagues went up by nearly 8%, they report today in Nature Human Behavior. That means that “police misconduct spreads … like a contagion,” writes Ojmarrh Mitchell, a criminologist at the University of South Florida in Tampa, in an accompanying editorial.
But Worden says the underlying complaint data make teasing out solid results impossible. “[The researchers] treat all complaints the same regardless of the source … or charge,” he says, noting that some complaints are unsubstantiated, whereas other instances of bad behavior may never make it into the database.
In addition, he doubts that experienced officers would be as susceptible to peer effects as their younger colleagues. “The socialization that officers undergo is more intense in their first year on the job. … By the time they’ve been on the job 3 to 5 years, they seem comfortable with their work.” So the authors’ argument, he says, “just doesn’t entirely ring true for me.”
Nevertheless, Stewart says the new method could be used to study all sorts of other questions about peer effects, including how political attitudes spread through social networks. But for now, he and Quispe-Torreblanca are focused on an upcoming visit to discuss their results with Scotland Yard: “I’m not sure there’s an earth-shattering result that will cause the police to change everything they do,” Stewart says. “But they’re all very keen to have no misconduct.”