Because most philosophies that frown on reproduction don't survive.

Wednesday, April 09, 2014

Testing in Daily Life: The Efficient Traffic Hypothesis

Josiah Neeley has an utterly charming article at The Federalist on his attempt to test his "efficient traffic hypothesis":
Probably the most frustrating part of my commute is when traffic in my lane is at a stand-still, and I’m forced to watch cars in the next lane zip by on their way to freedom. When you’re in that situation, the obvious thing to do is to switch to the lane that’s moving faster. But somehow it seems like as soon as you do, traffic in that lane grinds to a halt, while the lane you were just in starts moving again.

After brooding on this topic for longer than I care to admit, I developed a theory I like to call the Efficient Traffic Hypothesis (ETH). ... [which] says that you can’t beat the traffic. If you see a lane that’s moving faster than the one you’re in, chances are lots of other people have noticed the same thing, so it won’t be long before that lane slows to accommodate all the lane switchers. If it seems like whatever lane you’re in is always the slowest, that’s probably just a statistical illusion.

The Efficient Market Hypothesis is controversial, and there have been numerous studies trying to test whether it is actually true. The Efficient Traffic Hypothesis is less controversial, mainly because I just made it up. But being a seeker after truth, I decided to try to test it myself.

Using the website RandomizeMe, I created my own Randomized Controlled Trial (the same sort of test used to determine the efficacy of new pharmaceuticals). Every morning before leaving for work I flip a coin to determine whether that morning’s group will be in the “control group” (where I trying to reduce my commute time by switching lanes when it seems advantageous) or the “left group” (where I stay in the leftmost lane for as much of my commute as possible). I then record the time at the beginning and end of the commute, along with any unusual features of the day (e.g. was it raining?)

The hardest part of the experiment is just sticking to it. When it’s a left hand lane only day, it’s often quite difficult to keep to the plan when my lane is going forward at a crawl. But then I remind myself that this is for Science, and I soldier on. Perhaps more importantly, I’ve noticed that my subjective sense of how bad the traffic is on a particular day doesn’t necessarily line up with the objective data. On many a day I feel like my drive has gone on forever, only to find that it wasn’t any longer than on previous days where it felt like I was flying down the highway.

After six weeks of observations, I am still months away from having a large enough sample to really test the theory. But I got curious, and so ran some preliminary numbers. It turns out that, excluding Fridays (which tend to have much lighter traffic), my average commute time when I stick to a single lane is 58 minutes and 36 seconds. By contrast, when I try to beat the traffic by switching lanes, my commute is reduced to an average of 58 minutes and 30 seconds.

I'm actually surprised that it's so even, though given the way Josiah frames it, I find it believable. During the year or so I had a 70 mile commute across the Los Angeles basin (which took an hour and a half in and a two hours back) I became convinced that it was actually the opposite: that by the time you notice that another lane is going fast and manage to switch into it, you've probably caught the point right before the peak of the "market" for speed in that lane and it's about to slow down from too many people switching in, so if you switch lanes all the time it will actually take you longer. However, I never tested my theory as Josiah is doing, and I find his results quite believable.


Jenny said...

Oh, commuter traffic! My daily joy.

I'm not really surprised that there is not a big difference in travel time if all the lanes are through lanes. My daily dilemma is the quickest and most ethical way to handle a bottleneck where you lose half your lanes over the course of a mile or two.

I live in a town of about 30K where the vast majority commute north 20 or more miles to work. My own commute is 38 miles. There is one exit south of town, which I use, that dumps about a third of the traffic onto the two lane interstate. This traffic is absorbed fairly well.

Up the road a few miles, the road widens to three lanes and there is another exit that about another third uses. I cannot say how well the three lane interstate absorbs this traffic because immediately after the merge ramps disappear, the interstate narrows to two lanes again. Massive bottleneck here.

The two lanes continue for another mile or so up the road to the final exit, north of town, where the last third enters. Another massive bottleneck. The interstate continues with the two lane for less than a mile before it widens to four lanes and we all go on our merry way.

So to sum up: two lanes, three miles of three lanes with an exit, about two miles of two lanes with an exit, four lanes. Reverse on the way home. It is a giant cluster**** and someone real smart came up with this plan.

The ongoing dilemma is how should you drive this thing? Should you merge into the through lanes as soon as possible or should you drive the ending lanes until they are gone? This is a source of unending argument on the local FB page.

Jenny said...

And related to Josiah's post, I love that he actually ran an experiment to find the best way. Now he can drive however makes him feel better and know he isn't handicapping himself.

J. Ghigliotti said...

I'm not terrible surprised at the results noting my own commute times, but I also had a similar initial theory.

Although, I didn't do any empirical testing of the theory, I have noticed that my nerves are a lot better off not worrying about switching lanes and blindly believing the theory.

I'm curious what his long term results will be though.

Kate said...

This theory makes sense to me. I believe these sorts of questions are studied extensively by traffic engineers. I do remember reading that it is best to use both your brakes and gas sparingly when in congested traffic, merge early when a lane is ending ahead, and let other people merge ahead of you.

This guy seems to be having some fun applying principles of fluid dynamics to traffic patterns!

Jenny said...

"merge early when a lane is ending ahead, and let other people merge ahead of you."

It seems patently obvious to me that this would be the proper way to make the traffic as smooth as possible, but I can assure you this is *not* what happens.

Agellius said...

Thank you for this, it's utterly fascinating to me. I had noticed that the "slowpoke" car I thought I had left in the dust always seems to end up right next to me at the next light. It's good to have this impression backed up by data (insufficient sample size or no). Maybe it will help me to mellow out on the road. : )