The site FiveThirtyEight has a lengthy article on some of the studied connections between speed limits and road deaths, titled, Why the Rules of the Road aren't Enough to Prevent People from Dying. It looks at how speed limits are set, how long ago the methodology was created for setting speed limits, and wonders if some revamped matrix of road safety mandates could reduce traffic deaths.
The piece starts with a story of a cyclist who dies in New York City after being hit by a car, but one in which the driver was traveling "under the 30mph speed limit." That leads to a discussion of how speed limits are set, using the 85th percentile framework: engineers measure how fast traffic moves on a certain road and peg the speed limit within five miles per hour the speed used by 85 percent of traffic. That methodology is supposedly dated to 1964 and focused more on rural highways than urban roads which have cyclists and pedestrians in close proximity to cars, and the FiveThirtyEight piece believes "There are lots of problems with the study and its applications" because of that.
It cites a 2009 study from the American Journal of Public Health that attributed 12,545 more road deaths from 1995 to 2005 to higher speed limits after the national limit was abolished. But in 2013 the National Highway Transportation Safety Administration was putting out data that showed traffic fatalities have only gone up once between 2005 and 2013. The one odd year was 2012, attributed to the warm winter that got more people in cars; in 2011, fatalities were the lowest they'd been in 62 years. Overall, fatalities in the decade after 2005 were down 25 percent. Speed limits in that time have only gone up. Still, you'll find heaps of studies that take either side of the debate.
Yet the piece takes the angle that, "one might assume that we set speed limits with careful calculations aimed at maximizing safety," and after zeroing in on NYC's Vision Zero, gets comment from an advocacy group that, "We know that human beings are going to make mistakes, but we can design the system so that those mistakes won't be fatal." No perfectly non-fatal system has ever been created, ever, but as we've seen before, there are many ways to skew the data in that quest. Head to FiveThirtyEight to read the story.