By Marta Tellado, President & CEO, Consumer Reports
We live in an era dominated by data—a time when complex algorithms can use your personal information to determine anything from your credit limit to the cost of a ride service. When an airline changes its ticket prices immediately after you visit a competitor’s site? That’s an algorithm responding to new information it’s learned about you from your browsing data. While there is nothing inherently wrong with customizing costs based on factors that are actually relevant—after all, why shouldn’t life insurance premiums be higher for stunt pilots than for librarians?—there is something terribly wrong when hidden algorithms result in higher prices across the board for communities of color.
Sixteen months ago, the independent, nonprofit consumer rights organization, Consumer Reports, partnered with the journalists at ProPublica to launch a first-of-its-kind investigation into car insurance rates—another ordinary purchase with prices set by algorithms. We spent more than a year poring over reams of data in four states: California, Texas, Missouri, and Illinois. And what we found was startling.
In each state, we discovered that insurers were charging higher rates in some minority neighborhoods than could be explained away by reasonable factors. Within the same cities, drivers living in minority areas were often paying an average of 30 percent more than drivers living in white areas with similar levels of accident costs. In other words, after accounting for relevant factors that could explain price disparities between neighborhoods—such as the average costs of claims—major gaps still existed that don’t appear to have a legitimate rationale.
In Chicago, two equally safe drivers illustrate the consequences. Otis Nash, who works two jobs to support his wife and daughter, lives in predominantly-black East Garfield Park, while Ryan Hedges is an advertising executive in the more affluent—and mostly white—neighborhood of Lake View. Nash’s neighborhood has more poverty and a higher crime rate, but the data shows that it’s actually safer from an insurance standpoint: over the past few years, insurers have paid out 20 percent less for injury and damage claims in Nash’s zip code than in Hedges’. The two men have the same Geico insurance, and both are rated as good drivers. Hedges pays about $55 per month on his 2015 Audi SUV; Nash pays almost four times as much to insure his 2012 Honda Civic.
Stories like these shatter the conventional wisdom that auto insurers have used for decades to justify charging higher premiums in minority neighborhoods: namely, the claim that the risk of financial losses from accidents is greater there than it is in whiter, wealthier parts of town. We can’t say with certainty whether these findings are the result of intentional human discrimination or, more likely, algorithms with inadvertently discriminatory inputs. But we do know that insurers have often used factors that have nothing to do with a driver’s record—such as occupations and credit scores—to set rates that can disproportionately punish people of color.
We also know that, whatever the cause, the result is what lawyers call ‘disparate impact:’ a practice that, regardless of whether it is intentional or not, leads to unequal outcomes. When drivers like Otis Nash are forced to pay higher premiums, it becomes more difficult to afford a car, which in turn limits job prospects and makes it harder to get to work, which in turn makes it a struggle to pay rent, and so on down the line. When whole neighborhoods are subjected to higher prices, these ripple effects can be devastating—family budgets drain more quickly, employment opportunities are hindered, and communities are restrained from growth. It would be one thing if price disparities could be justified by legitimate risk, but we now know that they cannot. And when safe drivers with poor credit scores are charged more than dangerous drivers who happen to be wealthy (a practice our research has exposed in the past), vicious cycles are perpetuated in communities that have historically struggled.
Minority neighborhoods have for generations been preyed upon by industries conspiring to deny services or hike up their rates. The era of big data hasn’t solved that problem, but it has made it harder to assign responsibility to humans acting in bad faith. Our hope is that, by putting a spotlight on the issue of car insurance disparities, we can mobilize consumers to press for real solutions, such as state rules that increase pricing transparency and limit the demographic factors companies can use to set rates. But as more and more of the costs of living become entangled with algorithms and our personal data, we must remain vigilant to ensure that our values as consumers—including fairness, equality and transparency—are what’s driving the evolution of the marketplace.
