(Courtesy of Flickr user Miika Niemelä, under a Creative Commons license.)

Traffic fatalities in the United States have been plummeting for years, a major victory for regulation (strict drunken driving laws have helped) and auto innovation (we have safer cars). But that progress obscures a surprising type of inequality: The most disadvantaged are more likely — and have grown even more likely over time — to die in car crashes than people who are well-off.

New research by Sam Harper, Thomas J. Charters and Erin C. Strumpf, published in the American Journal of Epidemiology, finds that improvements in road safety since the 1990s haven't been evenly shared. The biggest declines in fatalities have occurred among the most educated. As for people 25 and older with less than a high school diploma, fatality rates have actually increased over time, bucking the national trend:

The underlying issue here is not that a college degree makes you a better driver. Rather, the least-educated tend to live with a lot of other conditions that can make getting around more dangerous. They own cars that are older and have lower crash-test ratings. Those with less education are also likely to earn less and to have the money for fancy safety features such as side airbags, automatic warnings and rear cameras.

The number of trauma centers, the researchers point out, has also declined in poor and rural communities, which could affect the health care people have access to after a collision. And poor places suffer from other conditions that can make the roads themselves less safe. In many cities, poor communities lack crosswalks over major roads. The residents who live there may have less political power to fight for design improvements like stop signs, sidewalks and speed bumps. As a result, pedestrian fatalities in particular are higher in poor communities.

"It's true that there are big differences in the quality of the residential environments that people have in terms of their risks of accidental death as pedestrians," Harper says.

The role of behavioral differences is murkier. Some studies show lower seat-belt use among the less-educated, but seat-belt use has also increased faster among that group over time, meaning socioeconomic differences there are narrowing. Data on alcohol use is also conflicting.

The chart above, based on National Center for Health Statistics data used by the researchers, captures miles traveled not just by car, but also bus or other motor vehicles (the poor are more likely to use transit, while the wealthy travel more by private car). The fatalities, though, also include the deaths of pedestrians and cyclists struck in car crashes.

In 1995, these death rates — adjusted for age, sex and race — were about 2.5 times higher for people at the bottom of the education spectrum than those at the top. By 2010, they were about 4.3 times higher. That means the inequality of traffic fatalities is getting worse, even as it looks nationwide as if our roads are getting safer.

As we increasingly fantasize about new technologies that will save us from our own driving errors — cars that will brake for us, or spot cyclists we can't see, or even take over all the navigation — we should anticipate that, at first, those benefits may mostly go to the rich.