Let me start by saying, I love Mythbusters. They've done some truly impressive things over the years; not to mention the mainstream-ification of science. But then maybe I say that because as a pop-science programme it plays right into my demographic. Perhaps people who know about all the other things feel like I feel when they test traffic "myths".
Now, I'm not talking the ones like "fuel consumption of A/C vs windows down" or "parcel delivery with only right-turns using less fuel", because these are perfectly valid things to test. No, I'm grumbling here about them testing things like "shockwave traffic jams" or "4-way stop vs roundabouts". These are actual things, you don't need to contrive circumstances for them. You can go to any highway and see them. We are beyond the realm of science and firmly planted in engineering. Though perhaps the US view of roundabouts does rest in the mythical, this is a different problem. The particular myth that I write this about here is "4-way stop vs roundabouts" which I have seen via a clip posted by James Snowdon on Facebook this morning. Thanks James.
So what's irked me this time then?
Overview of Test
Essentially the Mythbusters are testing the capacity difference between the 4-way stop (one of the most common basic US junctions) and a simple 4-arm roundabout. In both cases, the junction is the intersection of two figure-8 style arrangements so each driver comes back around to the junction once they leave it. This enables them to undertake the test with a limited number of drivers.
Experimental arrangement for the 4-way stop
Seems right, fine? Probably is. Could perhaps question how many drivers they have, but won't. The drivers are "randomly" directed one of three ways at the junction and circulate for 15 minutes. This is run twice after drivers receive some test time to become accustomed to the junction.
4-Way Stop in action
On average they found an equivalent hourly inflow of 1540pcu/h. Quite good actually from a stand perspective, considering there is no directional deference rule in the States and this was mixed turning where there was no left-turn waiting space that didn't block the junction. That being said, Americans are used to this sort of junction, we (the UK) would use a mini-roundabout and even then if all 4 arms are equally loaded, capacity is atrocious. In fact, we wouldn't use them for anything over a 500pcu/h total inflow (para 2.15). In such a case, we'd have to use a major/minor priority crossroads or, more likely, signals. So well done America on that one. For interest, they then test human control and find a fall of 30% capacity over the uncontrolled arrangement. Not surprising, but equally not where my irk lies.
They then rearrange the water-filled to a roundabout:
They found an equivalent hourly inflow of 1840pcu/h; some 20% more than a 4-way stop. What's wrong with that? Well everything.
ICD vs Roundabout Type
Extrapolating from the lane widths of the arms of the junction, vehicle sizes and ground markings, the diameter of the roundabout (the Inscribed Circle Diameter or ICD, to use the proper term) is, at best 28m. Now this value is a threshold because under it, you can't get an HGV around a central island. So for anything less than a 28m ICD, we use a mini-roundabout. Once a junction becomes a mini-roundabout, as drivers we are concerned with the approaching traffic on other arms and not simply the circulating flow. The capacity falls and they junction is a whole different beast.
Central Island and Layout Constraint
Let's give them the benefit of the doubt and say its just over 28m ICD. However, if we do that then by the same measurement, the central island is around 16m in diameter. When designing a roundabout we consult the design standards and in such a case we find that for a 28m ICD, you want to be using a central island of only 4m diameter. The reason for this is so HGVs can make the turn without driving over the island.
But "wait", I imagine I hear you cry. There aren't any HGVs here so why is that a problem. Well, either way, you put a overrun area around the island. Drivers circulate around the overrun area (on circulating carriageway roughly of the dimension shown) and if HGVs need to, they sweep over it. Whilst the HGVs don't matter here, there is a psychological component to the driver for being enclosed, called a layout constraint. In essence, the drivers here will feel like they're driving next to a wall, this will slow flow on the Mythbusters roundabout, reducing throughput.
The final geometry issue relates to the entry deflection, namely there isn't one. On approach to the roundabout, the test drivers arrive roughly perpendicular to the circulating carriageway. On a well designed roundabout, the entry would be deflected so some component of the turn onto the roundabout occurs before you get to the give way (of if you're American, 'yield') line. This is usually achieved by a combination of kerbside radius to lead you onto the roundabout, and a splitter island between your approach lane(s), the adjacent exit lane and the circulating carriageway. It looks a lot like this:
Entry deflection and splitter island on a roundabout approach
Now, maybe we can argue this is a "compact" roundabout and so geometric site constraint meant we couldn't possibly build such a deflection. That's partly reasonable, the roundabout in the picture above has an ICD of about 70m if I recall. But you still wouldn't not design in entry deflection. You'd lay hatching; you'd narrow the approach and build in deflection; you'd still put in a splitter island, even a non-kerbed one. This isn't explicitly a capacity thing, it's a safety one.
One of the reasons roundabouts are significantly safer than just about any other junction type is exactly this reason. Collisions that happen on a roundabout are usually limited to rear shunt and side-swipes. These are much less likely to result in serious injury than a head-on or T-bone you'll get on a 4-way. Most accidents occur at junctions and the UK has a huge number of roundabouts and just about the safest roads in the world. Draw your own conclusions.
If you stuck through the video, you might have seen this:
Where's the waiting traffic?
That's just one example screenshot; it happened regularly in what we saw. For considerable periods of time, there are no waiting vehicles on some arms. Apart from a perfect demonstration of relative efficiency for the same number of vehicles, you can't possibly call what was measured the capacity of the junction. If the junction were at capacity then there'd be vehicles waiting. So to then go on and say it's a 20% increase over 4-way stop is plain wrong. It's obviously more than that.
Fortunately, the Mythbusters conclude the "myth" of roundabouts being more efficient than 4-way stops as "confirmed" and suggest it is time for America to adopt them. Whilst I agree with the sentiment, unfortunately, to do so by only 20% comes from the flaws in their method. I just wish I still had access to ARCADY, then I could properly calculate what the capacity of the junction is.
As we've seen, the capacity of a like-for-like junction is much higher as a roundabout and therefore a better idea which will reduce congestion, pollution, and save time and money. However, on top of that, the case for roundabouts is a safety one (though not necessarily for cycles if that were a concern in the States) not simply throughput.
Good experiment? Myth: busted.