Skip to main content
How to Measure Traffic Congestion

On April 11, we ran an article arguing that bike lanes don’t cause traffic jams if you put them on the right streets. Herbie Huff, a researcher at the UCLA Institute of Transportation Studies, and Madeline Brozen, the program manager of the UCLA Luskin School of Public Affairs Complete Streets Initiative, wrote to us Tuesday, saying there was a methodological flaw in the analysis. Their letter is below, and you can read a response from the article’s writers, Aaron and Gretchen Johnson, below that. Both pieces have been lightly edited for clarity.

Here’s the letter by Huff and Brozen: went viral in our circles earlier this month, with a post entitled “Bike Lanes Don’t Cause Traffic Jams If You’re Smart About Where You Build Them.” Our colleagues and friends, people who love to ride bikes and want to make cities more bikeable, sent this article zooming around the tightly wound corner of the Internet that is all about #bikes. The post had a much-welcomed conclusion, backed by data, charts, and statistical tests: bike lanes have a negligible effect on congestion.

Read carefully, though, and the 538 analysis quickly falls apart — the authors measure congestion wrong. The article uses a volume-to-capacity ratio (V/C), the number of cars that go by a certain point on a road counted against that road’s theoretical capacity. But counting how many cars go by doesn’t measure congestion. To see why, think about a road suffering an intense traffic jam. In stop-and-go traffic, only a few cars would pass by a traffic counter over say, five minutes. Then think about that same road late at night, when the road is practically empty: the same low number of cars might pass by the same counter in a five-minute span. This makes it clear that congestion is really a measure of traffic density, not just volume. Taken statistically, these counts of traffic volumes don’t tell you anything about congestion.

By focusing on the volume-to-capacity ratio, rather than a measure of traffic speed or traffic density, the authors miss the effect of bike lanes — even when it’s staring them in the face. Just before their main argument, they write:

“[E]ach road seemed to have about the same traffic volume after its bike lane was installed. Running a statistical test … confirmed that there was no difference in [average daily traffic] before and after the installation of the bike lanes.”

Hold on. This is an incredibly powerful statement: These roadways were able to move the same number of cars and provide space for people on bikes. Reducing a roadway by one lane and achieving the same volume of cars means you’re doing more with less, not that the roadway is necessarily experiencing congestion. They reinforce this point later with an example from New York City, which directly measured travel times before and after the addition of bike lanes, and found that travel times didn’t change. Again, hold up — there was no change in travel time! This is what really matters when you talk about congestion.

What happened? Did some people switch from biking and driving? Did they make fewer trips during congested peak hours and more during the off-peak? Did some drivers switch routes? Focusing narrowly on statistical tests distracts from the true story of how places and people change as city infrastructure changes.

Riding a bike is simple, but transportation systems and the people that travel through them are complex. People are motivated in various ways for living, working, and getting around in the ways that we do, and congestion is just one factor among many. Data on all this can be hard to come by, and in particular there has been too little data and rigorous analysis about people on bikes. So we welcome’s entry into the conversation. And we hope future posts will grapple more seriously with the complex workings of both streets and people.

Here’s the response from the writers, Aaron and Gretchen Johnson:

We agree that measures of speed or the number of cars in a unit length of road at the peak hour would have been ideal to use. New York City measured this with the travel time, which is why we included it in our post. However, this information isn’t commonly available across a lot of cities and roads. What they have is the AADT [annual average daily traffic] or some measure of volume. We chose AADT, because the data was available and easily understandable to a broad audience. So, we were able to use that to compute the V/C ratio, which is one accepted measure of traffic used by planners and engineers, in part for precisely the same reasons. The theory (available in the Highway Capacity Manual) is that a certain V/C ratio on a certain type of road gives you a certain traffic speed and level of congestion (called “levels of service“). Those are the regions in the figures with “heavy congestion” or “severe congestion.” While finer data allowing for a more rigorous analysis would be nice, we feel confident that our analysis was the correct avenue to take for an article on FiveThirtyEight.

Lisa Chow was previously a features editor for FiveThirtyEight.