For years, police records have shown that black drivers tend to be less likely than white drivers to turn up with guns or drugs when searched at traffic stops. At the same time, black drivers are three times more likely than white drivers to be subjected to these searches, according to a 2013 federal survey.

You might object (or not) to the idea of racial profiling — officers taking race into account when determining whether to stop and scrutinize someone. But few would endorse racial profiling that clearly doesn't work. It's apparent, though, that ineffectual, biased policing continues in many places — more black people are stopped and searched for contraband, while more white people are found with it.

The police, it seems safe to assume, want to do their job most effectively. So the puzzle is: Why are they continuing to search drivers in an inefficient way?

The question gained new urgency after the New York Times published on Sunday a story that assembled records from four states that keep track of who gets searched at a traffic stop, and whether anything was found.

When drivers get pulled over, the police may request to look inside the car if they suspect there might be guns or drugs. Officers have considerable discretion regarding whom they ask to search.

This chart from the Times analysis shows how often black people were searched at traffic stops compared with whites. As you can see, the Connecticut State Police searched blacks 2.6 times as often as whites; the Rhode Island State Police searched blacks 2.5 times as often; the Illinois State Police searched twice as often; and the North Carolina State Police searched 1.5 times as often. In Chicago, blacks were searched more than five times as often as whites.

But all that extra scrutiny doesn't seem to be turning up any more illegal stuff. In Connecticut, Illinois and North Carolina, police found contraband less often when they searched black people than white people. In Chicago, for example, blacks were 30 percent less likely to be found with any illegal goods. State police officers from Rhode Island were the exception; they were more likely to search black drivers, but they also more likely to find contraband.

The implication is that some police officers in these states still harbor some — perhaps unconscious — bias against black drivers. It's a persistent problem, and one that even may be widening over time. But it's very perplexing.

In theory, if officers notice that their searches of black drivers turn up empty-handed at a disproportionate rate, they might second-guess themselves the next time they pull over a black driver.

A clue lies in how infrequent these searches are — and how infrequently police officers actually find anything. The federal survey from 2013 suggests that only about 3.5 percent of traffic stops end up in a search. The Times puts that number closer to 2.5 percent. Of those stops, the police turn up contraband perhaps only a quarter of the time.

And so one possibility is that the police just struggle to see the trend through all the noise — and thus have trouble overcoming any unconscious bias.

Take Chicago as an example, because Illinois makes this information easily available.

Between 2009 and 2013, Chicago police stopped over 200,000 white drivers, but only searched 906, of which 237 had contraband. Of those searched, that’s a hit rate of 26 percent. In that same time period, they stopped over 300,000 black drivers, searched 6,593 of them, and found contraband in 1,232 cases. That’s a hit rate of 19 percent.

So yes, the black hit rate in Chicago was about 70 percent the white hit rate. But in practice, the difference between 26 percent and 19 percent can be hard for officers to notice. When police are searching only a few out of every hundred cars they pull over, an individual officer might only witness a handful of positive hits in a year. That's not enough feedback to get a good sense of whether one's hunches about people are accurate, or racially tinged.

The Times analysis also highlights Hoffman Estates, a small suburb of Chicago, because the police there are prolific in the number of traffic stops they make. But police there, like the police in Chicago, are quite unlikely to actually search the cars they pull over. In 2009, Hoffman Estates officers pulled over 1,815 black drivers, searched seven — and found no contraband in any those cases. That same year, they pulled over over 15,000 white drivers, searched 40, and found 13 with contraband.

Recent experiments in psychology have shown what many already suspect: Racial bias is hard to extinguish from people’s minds. We expect our police officers not to let race affect their decision to search a car. But because only a tiny fraction of traffic stops result in actual contraband being found, individual officers might not get enough experience to learn whether they're conducting their searches in a biased way.

Even though the overall data show that officers err on side of being more suspicious of black people, it may be near impossible for any one officer to figure out if he or she is being too suspicious, or not enough.

To increase equity, it probably won't be enough just to tell police officers to search fewer black drivers. But education can help officers understand their own implicit biases.

Sometimes, it simply takes more experience. In laboratory experiments, people are much more likely to accidentally shoot unarmed black men in a video game asking them to make a split-second decision. But police officers who train extensively in these "shoot, don't shoot" scenarios show no bias at all.

The Times story cites a study from Northeastern University that credits new training measures for cutting down on racial disparities in traffic searches in Rhode Island. One popular curriculum used in police departments around the country encourages officers to reach out and have more conversations with minorities in their community. It emphasizes that racism can be subtle, and often subconscious.