Study: Drivers in Less-White Zip Codes Pay More for Auto Insurance

David Crockett, MBA, Ph.D.:

Insurance firms use ZIP codes as a proxy for a neighborhood. Postal codes are not an unreasonable proxy for a community, but they do come with some problems. For Black and Latinx consumers, when companies use ZIP codes, they are very likely to be incorporating several factors that probably have nothing to do with how good a person is as a driver or even their propensity to pay their bills on time.

A lot of African American neighborhoods are overpoliced. Many people understand that, but sometimes what I think gets missed is the way they are overpoliced. What we end up focusing on, and rightly so, are acts of violence. But the most common interaction that anyone has with law enforcement is traffic enforcement. When you live in a community that is overpoliced from the start, you know that you will have more interactions for traffic violations. Study after study makes this clear. It's not controversial to say that Black and brown people are ticketed more often for the same kind of behavior as white drivers.

Kalinda Ukanwa, MBA, Ph.D.:

I cannot definitively say what insurers are doing with ZIP codes since I do not have access to their risk assessment methods. What is likely happening is that auto insurers have algorithms that use ZIP codes as one of many predictors of an applicant's insurance risk. The algorithm is probably using information about the individual (e.g., driving record, credit score, vehicle, etc.) as well as the average behavior of groups he is a member of (e.g., residence ZIP code, gender, age, etc.) to predict risks for accidents, auto theft or other damages. ZIP codes may also correlate with the demographics of the residents living in the ZIP code, which in turn can be used to predict risk.

However, what is not clear is whether higher rates of theft, vandalism, accidents or other insurance risks in some ZIP codes are because there truly are more incidents in these ZIP codes or just more reporting to insurance companies of these incidents. For example, lower-income areas may have increased reporting of incidents to insurers because they do not have the ability to pay out of their own pockets for repairs. In higher-income areas, they have that ability, so residents may not report as often to insurance companies in order to avoid higher premiums down the road. A more accurate option for insurers is to look at information about the individual, such as driving record and driving habits, regardless of what other people in his ZIP code are doing.

Jerome D. Williams, Ph.D.:

It’s my understanding that insurers will look at various factors associated with each ZIP code to determine the appropriate rate for people living in that ZIP code who apply for insurance. Of course, rather than lump everyone in a particular ZIP code as being equally assessed for insurance premiums, I would argue that the individual’s driving record and other factors also should apply.

Vanessa G. Perry, Ph.D.:

The problem with using ZIP code to assess risk is that a ZIP code often serves as a proxy for race and ethnicity. So, although it is illegal to use race as a factor to determine insurance rates, companies can use ZIP codes instead and end up charging higher rates to minority consumers.

Douglas Heller:

Auto insurance companies use ZIP codes or, occasionally, other geographic boundaries such as census tract, as part of the formula to determine the premium charged to drivers. Typically, the premium is tied to the ZIP code in which the driver lives, regardless of where they actually do most of their driving. Still, there is some degree to which there will be differences in the riskiness of driving in different geographic regions. Jammed-up cities will have more low-speed accidents (high “frequency,” low “severity”), while wide-open rural roads will have fewer accidents per mile traveled, but they can be much more dangerous (low frequency, high severity). For the basic liability coverage everyone needs, using policyholders’ place of residence is meant to help insurers assess the likelihood that a resident of a region will cause an accident (the “frequency” of losses) and how costly those accidents will be (the “severity” of losses). Combine those two risk factors for a particular region, and insurance companies argue they can do a better job of determining the risk and setting the premium for each customer.

That might make sense and could be fair if a limited number of broad geographic swaths were used, but when small divisions like ZIP code or census tracts are used, problems arise. Not unlike the blatant red-lining of years past, research shows that ZIP codes with high proportions of people of color still face higher rates on average. What’s more, these high-premium ZIP codes are sometimes adjacent to much lower-priced ZIP codes that are whiter and wealthier. In my research, I’ve found that good drivers sometimes see premiums double simply because they live on the wrong side of a ZIP code border even though they may be across the street from a more affordable policy. Whether or not actuaries might reasonably argue in favor of drivers in one large region paying somewhat more or less than another region, there is no justification for a basic liability insurance policy from the same company costing one good driver 20%–30% more, let alone 100% more than their neighbor.

Insurance regulators in every state should analyze insurance prices by ZIP code and see how it matches with the racial and ethnic makeup of the communities. In most cases, it will be unnerving, if not surprising. They and state lawmakers should take steps to make sure insurance companies increase the weight given to things like a driver’s accident history and annual mileage and reduce the effect living in a particular ZIP code has on premiums.

David Crockett, MBA, Ph.D.:

Certainly they do, but these are things that are not connected to them being worse drivers or higher accident risks. Many of these things that we are talking about now are the residue of residential segregation. Many people would be astonished to know how high the level of residential segregation is in many cities across the U.S. As metropolitan areas become more diverse with a higher proportion of Black and brown people, residential segregation rises.

In the U.S., white people claim the most resource-rich space. This happens even among white people of modest means. You may be the poorest person on the block, but you live in an area with good schools and maintained roads. On the flip side, Black and brown households of even middle-class standards are often living in the worst kinds of neighborhoods. With this imbalance, you see what a lot of people describe as the "Black tax." Black and Latinx people tend to pay premiums for certain things that are a function of the neighborhood they live in that don't have anything to do with them personally.

It is my understanding that even after we control for things like driving record, credit score or vehicle type, so that we are comparing apples to apples — Sam Jones in one neighborhood and Shaquille Jones in another community — you still see racial differences in premiums. Even when comparing the same driving record, these things are largely a function of the legacy that we have left and are still living with; that is, residential segregation.

Kalinda Ukanwa, MBA, Ph.D.:

Credit score is most likely to have a disproportionate impact on people of color in terms of increasing premiums. Prior research well documents that people of color have lower credit scores, on average, compared to white counterparts, even when all other financial factors are equal. Many institutions use credit scores in their algorithms to predict risk, even if the risk they are predicting has nothing to do with credit risk. It would logically follow that systematically lower credit scores for people of color will translate into disproportionate increases in premiums.

Jerome D. Williams, Ph.D.:

I have no hard evidence that this is the case in the insurance industry, but I certainly have seen evidence of this in other sectors of the economy. I have been examining marketplace discrimination for over 40 years, and in our book “Consumer Equality: Race and the American Marketplace,” we offer an abundance of evidence that consumers of color are not treated equally in the marketplace. This evidence includes a disproportionate emphasis on consumers of color as shoplifters, when in fact, they engage in shoplifting percentage-wise significantly less than whites; disproportionate use of security tags in stores in neighborhoods that are predominantly minority compared to neighborhoods that are white; disproportionate higher prices for the same product in chain stores in minority neighborhoods compared to those same items in the chain stores in white neighborhoods; and food delivery services that refuse to deliver in certain neighborhoods that are primarily minority compared to neighborhoods that are primarily white. I could go on and provide numerous other examples. I would expect to find similar examples in the insurance industry primarily based on anecdotal evidence of ZIP codes where I have lived and the insurance rates where many of my white colleagues have lived.

Vanessa G. Perry, Ph.D.:

The use of credit scores is likely to have a negative disparate impact on Black and Latinx consumers. One reason is that members of these groups are significantly less likely to have a credit score in the system than white consumers (i.e., more likely to have no credit score on file).

Access to credit has historically been a challenge for consumers in minority and lower-income communities due to wealth effects as well as differential access to and higher costs for financial services. Despite gains in credit access and homeownership in the early to mid-2000s, minority households have been slow to recover from significant losses in home value and higher foreclosure rates experienced during the housing crisis. There is also concern about credit reporting during the pandemic, which has had a disproportionately negative impact on the financial well-being of Black and Latinx communities. These patterns are undoubtedly reflected in credit data and by extension existing credit scoring models. Thus, due to the effects of structural inequality, wealth and income disparities, discrimination over time, and the differential impact of major economic crises, average credit scores are lower for members of these groups.

Douglas Heller:

In addition to ZIP codes, insurance companies use various non-driving, individual characteristics that lead to higher rates for people of color and lower-income drivers, even when they have pristine driving records. These characteristics, or “rating factors,” include drivers’ occupation, educational attainment, homeownership status and credit score. Insurers insist that they never ask a customer their race, ethnicity or income level when they defend their pricing policies. But that just exposes the fact that companies either don’t understand or simply dismiss the systemic biases that persist in our society. Many of these non-driving, socio-economic factors insurance companies incorporate into their algorithms are proxies for race and income that tend to raise rates for low-income drivers and people of color. So, blue-collar renters with a high school diploma and a moderate credit score will pay significantly more for exactly the same coverage as a white-collar homeowner with a college degree and excellent credit, even if they have the same driving record, drive the same car and live in the same ZIP code.

These personal characteristics do not align perfectly with race and income. Still, federal demographic data are clear that African American and Latinx drivers are more likely to have blue-collar jobs and less education and have lower homeownership rates and lower average credit scores than white drivers. This, too, is a legacy of systemic racism. Testing reveals that when these high socio-economic status drivers cause accidents or are even convicted of drunk driving, they are often charged less than the perfect driver without the same social status. Because the same drivers who are penalized for these non-driving characteristics are also more likely to live in the ZIP codes that are charged the higher premiums, the pain of auto insurance discrimination is compounded, leaving more drivers in those same ZIP codes uninsured. If you’ve had a break in your insurance coverage, you’ll also face a higher premium in most states. It’s unrelenting.

Race and income are baked into car insurance pricing practices, and executives need to confront the harm done by the cumulative impact of these non-driving, socio-economic factors. Whether or not they are comfortable acknowledging it, the way insurance companies slice, and dice and price consumers disproportionately penalizes people of color, not only feeding off of structural racism but fostering it as well.

David Crockett, MBA, Ph.D.:

I want to stay in my lane because I am an inequality scholar, not a policy scholar. I do talk about policy in my work, but I do not know much about insurance policy. I can say in broader brushstrokes that a lot of this goes back to anti-discrimination policy in housing markets. It's the last thing that happened in the civil rights era, and it was the weakest kind of legislation. It caught a lot of resistance. Consider that many Black and LatinX neighborhoods are in cities that are underfunded and older inner-ring suburbs that are substantially overpoliced. I am very much in the camp of defunding the police and reallocating funds for things that contribute to the quality of life.

In many cities, Black and brown people get virtually all of the jaywalking tickets. Many of these places are infested with speeding traps. That is what those cities have to do to survive because they are underfunded from their state and federal governments. But these things raise a substantial proportion of tax revenue for many cities. But then, Black and brown people end up paying two or three times for this kind of disadvantage in increased insurance premiums. To change it, I am for strengthening anti-discrimination policy. I am less interested in seeing more people go to jail. I am more interested in policy that functions in specific ways. For example, companies should have to identify racial disparity. You have to find where racial inequality exists and do the analysis to understand underlying causes. Companies have no incentive to do that. An insurance company could say, "These things have nothing to do with us. We didn't cause any of these things to happen." It just so happens that while it disadvantages Black and brown consumers, the companies have made extra money. There has to be incentive to understand and analyze underlying causes and then to eliminate them.

Here's an anecdote of what can happen in policy and what everyday people can do. My colleague lives in a neighborhood that is several blocks of fabulous older homes. Those homes have been around for decades. They are all in pristine condition and great houses. A stone's throw from these houses is one of the poorest sections in our entire county. While we were in his home, he talked about being in negotiations with his insurance company over the premiums he was paying for being in that ZIP code. It took several rounds, but he was able to get his premium lower, based in part on what some other people in the area are paying.

Insurance companies would have you believe that their stuff is written down on tablets and carried from the mountain. Like, "This is what our algorithm said, what can we possibly do?" But all of these things are negotiable. This was not just a question of my colleague's negotiating skill, though I am sure he is skilled. In addition to these kinds of premiums, we commonly see that companies can be more attentive to white people's complaints. This is a common fact of doing business in the United States. I am working with a set of retailers right now, on these very issues, revisiting their policies. Companies must be made to be vigilant about these issues.

Kalinda Ukanwa, MBA, Ph.D.:

Legislation that restricts the use of data about protected classes like race and ethnicity would be important. This also goes for restricted use of data that proxies for protected classes, such as ZIP codes. However, legislation should also include the requirement that companies collect data about protected classes (but not use them in their decisions) in order to enable audits of their own service decisions. This would help to detect systemic disparate impact on consumers in protected groups. Furthermore, legislation that includes a requirement for transparency in the algorithms used to make service decisions like insurance premiums would be needed. The net effect to consumers would be fairer insurance rates.

Jerome D. Williams, Ph.D.:

It would work the same way as legislation to prevent redlining in the insurance and banking industries. Firms in these industries, when they employed this practice, would decide that they were not going to serve certain neighborhoods if they were composed primarily of ethnic-minority households, regardless of their creditworthiness or insurability. Consequently, they would, for example, take a map of the area in question and, using a red pencil, place a red line around the areas they were not going to serve and use such a marked-up map (often prominently displayed in offices in firms that employed this practice) as a visual aid to guide their responses to all future requests for loans and insurance policies from the area in question. Hence the term redlining. This practice was outlawed by the federal government through the Fair Housing Act of 1968. Consequently, it is now not as common a practice as it used to be, although it still figures from time to time in sporadic lawsuits and settlements.

Vanessa G. Perry, Ph.D.:

Legislation to limit pricing discrimination should require disparate impact testing, especially in the use of algorithms and artificial intelligence for pricing. One example of this is a 2013 regulation that articulates the standards for proving systemic discrimination under the Fair Housing Act.
According to the disparate impact (also known as discriminatory effects) standard, banks are liable for underwriting and pricing discrimination if there is evidence of significant outcome disparities that have no necessary business justification.

This liability does not require proof that the bank intended to treat one group of customers worse than another group. The same kind of regulation should be applied in the auto insurance industry. The industry could regulate itself by undergoing rigorous testing to ensure that its models and practices do not negatively impact historically disadvantaged groups. Further innovation and exploration in the age of big data would likely identify less discriminatory alternatives.

Douglas Heller:

The key to addressing both conscious and unconscious pricing discrimination is attacking both forms. Legislators around the country should approve laws that prohibit insurers from using non-driving, socio-economic factors, such as credit score, educational attainment and lapses in prior insurance coverage for rating and underwriting decisions. These are conscious practices of insurers that should not be used to price a product that state laws require everyone to purchase. Some states are considering such legislation right now. Rep. Tlaib of Michigan, Rep. Watson-Coleman of New Jersey and Sen. Booker of New Jersey have proposed federal legislation in this vein that would set a national standard for anti-discriminatory pricing.

Regulators have an important role, especially with the increasing use of big data in the insurance industry. State insurance commissioners should apply disparate impact tests to all companies’ pricing algorithms to identify and remove the unconscious and conscious bias from pricing. As insurers often explain, insurance is built on discrimination—between, for example, good drivers and bad drivers and between occasional drivers and those with long commutes. However, the job of regulators is to disrupt and stop unfair discrimination, which they are authorized to do under state laws throughout the country.

One place to start is with the approach that California voters took when they enacted Proposition 103 back in 1988. Under that law, premiums must be based first on driving record, second on miles driven annually and third on years of driving experience, with other approved factors having less impact. Socio-economic proxies such as credit history and prior insurance coverage are prohibited. When insurance companies remove these unfairly discriminatory pricing mechanisms, they must emphasize factors more relevant to risk, which means there will be a restructuring of prices such that drivers with poor safety records must bear their share of the cost. In contrast, safe drivers of all backgrounds see savings.

Another option for states is creating a low-cost insurance program for low-income safe drivers; California has such a program where policies are less than $500 a year. More about that program and how it might be applied in other states is in a paper I wrote for Maryland’s Abell Foundation in 2019.

David Crockett, MBA, Ph.D.:

Shop around. Insurance is one of those things where people can quote you radically different premiums for the same product. Research what you should be paying for your area and negotiate what you can. People must look to see what is going on in their areas around issues like this. One of my favorite organizations that I just became aware of since I have been doing this kind of work is a tiny voluntary association called the Funerals Consumer Alliance. That is a service that most of us do not want to think about. Consequently, people pay radically different prices for the same kind of service. It happens all the time. One of the things members do in the cities they are located in is price out a standard funeral. They print out a booklet and make that available. You can take action as an individual shopper, but we have a whole history in this country of consumer activism where people get together for some of these things we can't do on our own. Your negotiating skills are not better than the insurance company's. But maybe you can get together with other people facing the same kind of issues, and make some change.

Kalinda Ukanwa, MBA, Ph.D.:

Contact your insurer to ask for details on how they arrived at the rate. Ask them what the rate would be with your profile if you lived in a different ZIP code. This may open the way to renegotiating your rate. Also, know your rights based on the state that you live in. For example, California proposition 103 restricts insurers from using ZIP codes in insurance decisions. Finally, consumers may want to look at insurance providers, such as Root Insurance, that offer telematics programs. Telematics programs use a small device installed in your car to report in real time on driving performance. In fact, most major insurers now have telematics programs. These programs use driving history and car usage to set insurance rates instead of predictors like ZIP code and credit score.

Jerome D. Williams, Ph.D.:

I’m not sure how much recourse an individual driver may have legally under current laws. Just as consumer rights groups have gone to court to eradicate redlining in the banking industry, similar efforts may be needed to eradicate the practice of retail redlining in the insurance industry.

Vanessa G. Perry, Ph.D.:

Residents of ZIP codes who are concerned about their premiums should contact their insurer to request detailed documentation—many of these disclosures are required by law. They can also file a claim with their state insurance regulatory agency.

David Crockett, MBA, Ph.D.:

This notion of the "Black tax" appears in the cost of everything. It begins to overlap with the criminal justice system in the form of increased parking tickets, traffic tickets and jaywalking tickets. It provides entrée into the criminal justice system and mass incarceration for a lot of people. They arrive there through unpaid parking tickets and speeding violations. Then, they end up driving without a license or insurance or both. This adds up.

Remember the gentleman killed before George Floyd, Philando Castile. There were dozens and dozens of interactions with the police. Ticket after ticket after ticket, but no one drives that poorly. This is a function of the neighborhood. Beyond that, I grew up in North St. Louis. My parents lived a stone's throw from Ferguson, where Mike Brown was killed. It was a feature of everyday life. There are tons of tiny municipalities that make up St. Louis county, and many of them only exist to be speed traps. That is the only reason it's an administrative unit. They don't provide services to citizens. All they do is ticket people. It can ruin your life.

Years before Mike Brown, a guy played basketball with some of his friends at an outdoor court in Ferguson. He was sitting in his car, minding his business. The guy was perfectly legal, no problem sitting there, but he was approached by a police officer, who decided to arrest him and have him charged for making a false statement to a police officer because he reported his name as Mike and not Michael, which was on his driver's license. He was a federal contractor. He lost the federal contract. He lost his license to do federal contracting because you don't have to be convicted to lose your license; you just have to be charged. That man's life was ruined over nothing. This is the stuff that contributes to the racial wealth gap.

Right now, in the U.S., white people have 10 times the net worth of other groups. In Los Angeles, the median white household net worth is about $350,000. That's a pretty significant number. The median net worth of an African American household in Los Angeles is $3,500. The median net worth of Mexican-ancestry households in Los Angeles is $4,000. In Boston, the median net worth of African American households is functionally zero. That means half of the people in the distribution have a negative net worth. These things contribute to the cost that you pay for mortgages. They contribute to being in the criminal justice system and to the rates you pay on credit cards. If you live in a "white" neighborhood and you are just as reckless on the road, it is a near certainty that you will have fewer interactions with the police. You will pay a lower interest rate on a credit card, even when we equalize everything. These small little premiums, sometimes not so small, fit together to create a story of insane inequality.

Kalinda Ukanwa, MBA, Ph.D.:

It has the impact of taking more dollars out of the pockets of people of color. This leaves less for other financial needs in the household. Less available dollars for other financial needs such as food, rent, transportation costs and education means people of color will not be able to purchase as many resources, dollar-for-dollar, as their white counterparts with the same income. This perpetuates poverty and slows social advancement for people of color in the long run.

Jerome D. Williams, Ph.D.:

There’s no question that this has a negative disproportionate effect on the financial well-being of people of color. For example, if you take a Black family and white family, each with similar household incomes, the Black family will end up with less discretionary income, because they are spending more to purchase the same product because of where they live, even if they have a better driving record. In essence, this is perpetuating the cycle of poverty among people of color.

Vanessa G. Perry, Ph.D.:

Many studies have found that Black and Latinx consumers pay higher costs for financial services and higher transaction costs than similarly situated whites. These higher costs occur because of cumulative disadvantage (structural inequality, historical discrimination, and ineffective public policies) and persistent neighborhood racial segregation. Consumers who face higher payments and have fewer resources are more likely to incur late payments, and as a result, credit scores are lower on average for minorities relative to similarly situated whites.

Many metropolitan areas remain highly segregated by race and ethnicity. There is considerable evidence from prior research that home values in predominantly Black or Latinx neighborhoods tend to be lower than values in similarly-situated neighborhoods with lower minority concentrations.

Research also suggests that neighborhood housing appreciation declines significantly as the share of minority residents increases. These patterns are likely due to segregation and lingering effects of redlining. Using zip code as a variable in insurance pricing perpetuates these effects, and creates a negative, disparate impact on minorities.

Once these disparities get passed down to new generations of Black and Latinx Americans, even with higher levels of educational attainment, employment opportunities and access to economic and social capital, because they started at a deficit, it is difficult if not impossible to catch up to similarly-situated others. Thus, higher insurance rates are both caused by and purveyors of the large disparities in wealth between Black and Latinx households relative to whites.

David Crockett, MBA, Ph.D.:

We are in what one of my advisors used to call a "pregnant moment." There are moments in history that you know are going to be important. This is one of those. I hope we don't squander it. A lot of organizations and people in those organizations are sincere. They know there are issues. They know their organization has been part of the problem. How can we fix this? What can we do? They are beginning to look for answers.

I am working with a set of companies right now. One of my co-ops told me that decades ago, he was one of a handful of Black employees at PepsiCo when they took over the Aunt Jemima brand. There has long been opposition to that as an icon in Black America and inside the company. For years, people tried to convince the company that they needed to retire the brand. And that finally happened recently. Things look intractable until they are not. There was a sociologist named William Gamson who said, "Activists rarely understand the probability of success," and on the whole, that is a pretty good thing.

Kalinda Ukanwa, MBA, Ph.D.:

Algorithmic bias in service decisions such as insurance premiums can have not only long-term negative financial impact on consumers, but can also have a long-term negative financial impact on the firm. So many firms use algorithms to make service decisions because they believe the algorithms improve profits. My research suggests otherwise under certain conditions. My paper, "Algorithmic Discrimination in Service," shows that if firms used biased algorithms in service decisions, this can lead to reduced long-term demand and profits because of the word-of-mouth it generates. What is surprising is that the word-of-mouth doesn't even have to be negative, yet the firms can still lose.