More on the center median debate between citizens and engineers

David Cox says today’s Road Worrier column ("Median debate takes a turn") gave short shrift to the work done by him and some North Raleigh friends to critique a 2005 NC State University safety study comparing raised medians to center turn lanes on busy four-lane roads.

So I’ll provide much more space for the debate here.

Joe Hummer, an NCSU professor and traffic engineer, headed the study and interpreted its findings to say that roads with raised medians have fewer crashes than those with center turn lanes.   But Cox and a handful of his friends, who are not professional engineers, drew different conclusions from their examination of Hummer’s study and their own look at crash histories on some other roads.

Tim Niles, a retired computer programmer, said of today’s column:

I found the article to be biased, especially knowing the data that Siceloff had access to. Yes, he mentioned data from both sides of the debate.   But, he did it in a way as to give more weight to one side over the other.

And Cox, a computer scientist, said:

The reason we examined these studies in the first place was to come to some conclusions based on facts.   We felt that this was a better approach than simply complaining. I think it would be good to present the details of the studies that the DOT refers to as well as our findings on Brier Creek, Strickland, and other roads to allow readers to reach an unbiased conclusion.

It’s true that I left out lots of the discussion from both sides -- including detailed arguments offered by Cox and Niles, and details of countering views from Hummer and Jim Dunlop, a DOT traffic engineer.

Some of this stuff just wasn’t relevant.   Cox raised an alarming issue about “cross-over” crashes on roads with medians, but the evidence he cited came from a study about medians on high-speed freeways.   Not on a road like Falls of Neuse with stoplights, driveways and much lower traffic speeds -- quite a different creature. Not what we’re talking about here.

Plenty of the discussion was indeed pertinent. But there just wasn’t room in a 600-word newspaper column to get into what the study said about something, how each side interpreted it, and further comments from Niles, Cox, Hummer and Dunlop.

Cox and Niles asked me to give readers access to more of these details. So here goes.

Linked at the bottom of this blog post are copies of:

1) the 2005 study by Hummer (or you can click here for a powerpoint distillation) and
2) a 2008 critique by Cox, Niles and others on behalf of the North Raleigh Coalition of Homeowners Associations (NORCHOA).

And before we get there I will paste in excerpts from emails I got in the last couple of weeks:

from Cox, discussing his group’s study of crashes on other Raleigh roads, and
from Hummer.

First, from David Cox (5/12/2011):

Bruce,

In 2008 our group, NORCHOA, looked at the safety of medians vs. center turn lanes in regards to Falls of Neuse Road. At the time we were divided (pardon the pun) about which was better.   With an adequate number of signalized intersections, landscaped medians could look nice and be an asset to an area while providing sufficient entry and exit to neighborhoods.   On the other hand, are they really safer than center turn lanes?   To answer this question, we reviewed the literature on the subject – including the studies performed by Prof. Hummer – and reviewed accident statistics for local roads.   We came away convinced that center turn lanes are the safer choice.

Medians attempt to address two major classes of accidents: cross-overs and those involving turns to and from side streets.   One would think that medians act as a barrier to prevent cross-over   accidents.   However, we found this to be a misconception. According to Public Roads (a publication of the Federal Highway Administration, Vol. 68, No. 4 Jan/Feb 2005), “Cross-median crashes pose a significant hazard for motorists across the country, claiming numerous lives and causing millions of dollars of damage each year.”   The article states that there is no pattern to these crashes and causes “include everything from fatigue and improper lane changes to inattention and medical emergencies.”

The other class of accidents are those involving turns to and from side streets.   In theory, a full movement intersection has 16 possible contact points or places where an accident could occur.   At intersections where left turns are blocked by a median, there are (again, in theory) only 3 contact points.   This would lead one to believe that a median divided highway is five times safer than a center turn lane.

However, we found several situations where this theory doesn't hold up in practice.   Using roads suggested to us by DOT, we compared actual accident data from the city of Raleigh. We obtained three years of accident data (Jan 2005 through Dec 2007) for Brier Creek Parkway (built with a median) and Strickland Road (built without a median).   The data were obtained for a period of time before Brier Creek plaza was completely built out.   During this time Brier Creek Parkway actually had less traffic than Strickland.   We found the following:

Brier Creek Parkway was shorter, had less traffic, and less exposure but 10 times the crash rate.

Brier Creek had only 3 access points for a left turn.   Strickland had 30.   Yet, Brier Creek still had 7 times the crash rate for left turns.   This is peculiar given that left turn crashes are specifically what the median is supposed to limit.

There were only 3 accidents in a three year period on Strickland Road for left turns entering the road.   There were seven on Brier Creek.   And, remember, the Brier Creek segment is shorter and had only 3 places where a left turn could even be made compared to 30 opportunities on Strickland Road.

Strickland Road is more residential, you say?   Not quite.   Strickland road is about 50% business and 50% residential.   Strickland Road has a shopping center, 2 churches, a school, 2 office complexes, a bank and a day care center.   Strickland is not a sleepy little residential street.   In fact, it had more traffic per day during that three year period than Brier Creek Parkway:   15,000 vehicles per day compared to 10,000.

The same comparisons were done using two other local Raleigh roads:   New Falls of Neuse Road and Wakefield Pines where both have medians.   For the same three year period, New Falls of Neuse Road had a crash rate two times higher than Strickland.   Wakefield Pines had a crash rate three times higher than Strickland.   So, in each of these cases, Strickland Road, with a center turn lane, was considerably safer than the other similar roads in the area that were built with limited access medians.

Finally from Joe Hummer (5/18/2011):

Bruce,

Ni hao from Nanjing, China, where I am doing some lecturing.   I do not have access to my files and have only spotty internet connections, so my response will be just from memory.   Hope it helps nonetheless.

The main criticism of my work seems to be that there is a wide scatter in the data.   This is a weak criticism.   Of course there is scatter; collisions are very difficult things to predict and all collision prediction models have lots of scatter.   That does not invalidate the conclusion, which is that ON AVERAGE medians will prevent more collisions than two-way left turn lanes, all else being equal.   Of course there will be some cases when the reverse is true, but since we cannot predict those cases ahead of time they are not relevant to designers.   Over the long run, over a number of sites, medians will save collisions.

The criticism that medians only save a collision or so per mile per year is also baseless.   Those collisions add up over distances or over the years.   For some perspective, remember that on average a mile of road in NC has about 2.5 reported collisions per year.   Saving one collision when only 2.5 is expected suddenly seems pretty impressive.

Medians do not cost much more than TWLTLs [two-way left-turn lanes], and bring a number of other benefits besides safety, so from a cost-effectiveness point of view even a "modest" savings of one collision per mile per year is a great investment.

The main thing my study did was allow a comparison of apples to apples:   roads with exactly the same conditions except for the median treatment.   Many other studies do not do this, comparing roads with many other differences besides median treatment.   Comparing Briar Creek to Strickland, for example, proves nothing, as there are many other differences between those roads that could have caused any difference in their collision frequencies.   Briar Creek had some notoriously bad median openings for example.

I have great confidence in my study results because they are in concert with worldwide professional opinion.   Every study done with a decent methodology on this subject that I have seen comes to the same conclusion, that medians save collisions.   I was surprised with my study that the difference was not larger, but it was there and consistent in direction with the other studies.   There is very little certainty in traffic engineering as you know.   That medians generally save collisions when compared to the same roads with TWLTLs is one of our few certainties.   By the way, even in a crowded country like China all important roads have medians.

Documents: