For all the data analysis conducted by The Courant, no other topic comes close to generating the reader reaction we receive when looking at apparent racial disparities in policing.
That was evident once again this week with the release of fresh police-stop data showing – as previous releases have – that statewide, black and Hispanic motorists are stopped and ticketed at higher rates than white drivers.
That led to a flurry of comments on The Courant’s website – more than 100 at last count – most of which took exception to any implication that police officers might be treating minority drivers more harshly. Some were simply self-disproving diatribes – posts that used overtly racist slurs in arguing that racism was non-existent. But others took aim at the statistical methodology applied, some raising legitimate points, others misinterpreting or making incorrect assumptions about the analysis applied.
So, as we have in the past, here’s a primer on the data collected by the Connecticut Racial Profiling Prohibition Project and on the Courant’s analysis of post-stop behavior, along with responses to the most frequent issues that are raised whenever we dig into this data.
First, some history: Connecticut has been gathering police-stop data since 2000 – or at least departments were supposed to have been compiling the information, and the state was supposed to be analyzing that data. But most police departments ignored the law and more than a decade passed with no rigorous analysis. So in 2012, the legislature strengthened the statute, improving the form officers fill out with every traffic stop and turning the task of analyzing the data over to the Office of Policy and Management.
OPM, in turn, hired the Institute for Municipal and Regional Policy at Central Connecticut State University to make sense of the numbers. The Institute, led by respected researchers, applied a variety of statistical models to the data in an effort to identify individual departments in which an unexpectedly high percentage of blacks or Hispanics are being stopped. Racial-profiling studies have evolved over the decades, and researchers acknowledge, for example, that little can be determined by merely comparing who gets pulled over in a particular town to the demographic breakdown of that town. Similarly, it is not always meaningful to compare traffic stops to the estimated driving population in a town, because those estimates are imperfect and there may be actual differences in the rates at which different groups violate traffic offenses.
So scientists have looked for better methods, including trying to see if there are differences in the rates at which minority drivers are stopped during daylight hours compared to nighttime hours when the race and ethnicity of motorists is harder to discern before a stop is made. That presents its own challenges, however, as the driving populations at different times of day could be significantly different. So one analysis performed by the CCSU researchers looks at stop rates at the same time of day – but at different times of the year, when the earth’s position relative to the sun affects whether it’s light out or dark out. That method, known as the Veil of Darkness Test, is further explained here.
It’s a clever strategy. But given the complexity of developing a methodology to determine whether certain drivers are being stopped at higher-than-expected rates, The Courant has consistently pursued a different analysis, looking at what happens after an officer has identified a traffic violation and pulled over an offender. That’s the analysis that appears in Wednesday’s paper, and which we’ve done before with earlier releases of data. Each time we’ve look at the numbers, we’ve found that statewide, black and Hispanic motorists were more likely to receive a ticket rather than a warning, compared to white drivers pulled over for committing the same offense.
Wednesday’s story reported, for example, that among drivers pulled over for running a stop sign during the six months ending March 2015, 27 percent of white offenders were given a ticket, compared to 30 percent of black offenders and 39 percent of Hispanic offenders.
Some see statistics like that as evidence that non-white drivers are treated unfairly – even if subconsciously – by police. Others insist the data is garbage and resist any suggestion that there is broad-based discrimination by police, citing a variety of reasons why the Courant’s report is, as one commenter wrote, “another BS story.”
So let’s take a deeper look at what those reacting to the story are saying:
‘Blacks and Hispanics receive more tickets because they violate driving laws more often.’
“When you speed, you get pulled over. Just so happens that minorities do more of the speeding,” one poster wrote. Added another: “Guess what.Obey the law.If your not speeding or breaking a law. You won’t get stopped.Maybe the minority drivers don’t get it.” And a third wrote: “So this data just confirms that blacks and Hispanics need to obey laws like whites and Asians to stay out of trouble with the law.”
But The Courant’s analysis is not affected by variations in lawbreaking, if they exist. We analyzed only the outcome of a traffic stop after an officer has identified a traffic violation. So essentially we were asking: For every 1,000 white drivers, 1,000 black drivers and 1,000 Hispanic drivers who get pulled over for committing a particular offense, how many in each group get a ticket and how many get off with a warning?
‘The results are skewed by small numbers.’
With 18 months of data now available – more than 840,000 records – the statewide figures are highly statistically significant. The numbers are smaller when broken down by offense and town, but all of the numbers used by the Courant are significant at the 95-percent confidence level – the gold standard in research – and most are statistically significant at the 99-percent level.
‘The results are skewed by location.’
“If you have a city of 100K people and half are one race, one quarter another and the rest belonging to yet another, it is likely the the largest group will be pulled over the most and will receive the most violations,” one poster wrote. Added another: “In what areas was this data collected? Stop making bad behavior about race!!! Don’t want a ticket? Obey the traffic laws.”
Those posts repeat a frequent misunderstanding about the police-stop data. The Courant’s analysis only compares the rates at which different groups are ticketed. The population breakdown is relevant only in so far as a very small population could yield results that are not statistically meaningful. But as noted, all of the statistics cited by the Courant are based on numbers large enough to reach a high level of statistical significance.
‘What about that one study…?’
“Another BS story,” one commenter wrote, “this was debunked years and years ago in NJ. They put speed cameras on the [New Jersey Turnpike] and found, lo and behold, that minorities drove faster than whites. Ergo they get stopped more, and the impartial eye of the camera proved the point.”
Several posters made a similar point. But the New Jersey study, while important, doesn’t debunk the Courant’s analysis – or anything else. The Turnpike study was designed to measure the racial and ethnic breakdown of those driving on the Turnpike, and – for the first time – the racial and ethnic breakdown of those motorists who were speeding (based not on ticketing, but on actual radar checks on a representative sample of cars). The study was commissioned to address data showing minority drivers were pulled over at above-average rates. But the report, published in 2005, found that the percentage of black motorists traveling 80 miles per hour or faster was fairly close to the percentage of black motorists stopped for speeding and other moving violations on the Turnpike.
When the bar was adjusted lower than 80mph, the gap between speeding and stops reappeared, with black drivers pulled over at higher rates than they were actually speeding. But if threshold for pulling over a speeder were 80mph or close to it, that could explain the apparent “over-ticketing” of black drivers on the Turnpike.
But one study on one roadway does not establish that racial profiling does or does not exist anywhere in America. More importantly, the Courant’s analysis, as noted above, was not focused on evaluating who gets pulled over, but only what happens after drivers are stopped. Statewide in Connecticut, for all drivers caught speeding, 43 percent of white offenders received a speeding ticket, compared to 56 percent of black speeders and 49 percent of Hispanic speeders. No study has been conducted in Connecticut cities and towns to determine if black and Hispanic drivers as a group exceed the speed limit by significantly higher rates than white drivers, or if that could explain the large disparity in which speeders get ticketed and which don’t.
‘Police are not biased. Other, legitimate factors explain the disparity in ticketing.’
That could be. As we’ve noted before, correlation does not equal causation, and statistics cannot prove intent. So researchers look for “confounders” – outside factors that might influence a study’s findings and lead to an incorrect conclusion. As an example: men and younger people tend to be ticketed more than women and older people. Hispanic drivers stopped by police in Connecticut were younger than average and more likely to be male. Those factors could increase the percentage of Hispanic motorists who come away from a police stop with a ticket – for reasons unrelated to their ethnicity. (In reality, that may have a small effect on the data, but the gap in ticketing between white and Hispanic drivers is too large to be fully explained by differences in age and gender.)
But there could be other factors – and readers frequently cite one. “Did the study include any statistics about whether the driver gave the police officer an ‘attitude’? ‘Yes sir, no sir, I’m sorry for doing it, sir’ helps turn a ticket into a warning,” one commenter wrote. Another was far blunter, declaring with a broad brush: “minorities … do not respect law enforcement, and thus get belligerent when stopped by police.”
Does the demeanor of a motorist determine whether a police encounter is merely an unnerving experience or a $200 cash drain? That is not something traffic-stop data can help answer. Police leaders insist the attitude of drivers does not affect ticketing decisions, but if it does, that may raise entirely new questions about the fairness and propriety of police decisions.
Race remains a difficult subject in America, spawning strong and emotional reactions across the spectrum. “Another RACE BAITED article by the Liberal Courant stirring up the hornets nest, one sided,” one frequent poster wrote, objecting to Wednesday’s article. “If you break the law , police can pull you over. Solution? Obey all laws, go about your day , smile a lot.”
Serious social scientists know the reality is not nearly so simplistic.