Collision Statistics

I was passed along this article today, reporting on a sharp increase in pedestrian fatalities in Minnesota. In the StarTribune letter of the day yesterday, written by Michael D. Hoy, he questioned the conclusions of the reporter, showing that the fatalities fall in line with what is expected of normal fluctuations.

About two-thirds of the time, the statistic will lie within one standard deviation of the mean. This is what happened in the last 11 years. Two-thirds of the years had a death count between 33 and 45.

Also, about 95 percent of the time the statistic will lie within two standard deviations of the mean. Again, this is just about what happened. – Michael D. Hoy

It’s easy to look at collision fluctuations and be immediately concerned, but it’s important to remember that fluctuations can be random without any significant cause. That’s why it’s especially important to have good collision data collection methods in place when looking at collision rates and the safety of intersections. Sadly, most municipal resources are too scarce to spend a lot of time and effort on very involved collision reports.

If you’re a municipality or state agency focused on safety, you face a serious challenge everyday: with limited data, dollars, and man-hours available, how do you pick the most dangerous sites in your jurisdiction?

A background in statistics is always helpful, especially for researchers in this field. But for engineers who want to apply models everyday, there are plenty of methods already figured out for you to use. I highly recommend pursuing a graduate level course on highway safety like I did at NC State. Without delving into a statistics degree, you can pick up many effective methodologies to rank sites. Use them to your advantage and get the most efficient use out of precious tax-dollars while saving as many lives as you can!

It’s important to keep these things in mind:

Look at available data carefully. How long has the intersection been open? How much data do you have? Without enough historical data, it will be impossible to see any kind of trends.

Don’t revisit the same sites every year. Did you check the site last year? What were the findings? Unless the site has changed significantly, there’s no reason to check it again. There are tons of dangerous intersections that you haven’t investigated, get to those next.

Volume bias. Sites with the most collisions aren’t necessarily the most dangerous sites. Ranking sites by the number of collisions gives bias to the busiest intersections with the highest volumes. Similarly, sites at the bottom of the list may be dangerous, but have so few vehicles traveled on that they are skipped altogether.

An increase in collisions doesn’t imply a decrease in safety at the site. Especially in areas of high growth, increased collisions may just be because an increase in traffic. Check historical AADTs when you check historical traffic reports, and pick good comparison sites when comparing increases in collisions at one intersection with other intersections in your systems.

And try not to be overwhelmed by media.

News is always more interesting when there is a fright element, but there is no reason to believe that walking is any worse now than before.

Comments are closed.