The Washington, DC region is great >> and it can be greater.


Bus stop density correlates with speed

This graph compares the scheduled average speed by route with the average number of bus stops per mile for all bus routes in the WMATA published data.

The overall trend is for buses with fewer stops per mile to have higher average travel speeds. Buses that use grade-separated routes, like the 5A and other freeway buses, tend to have both fewer stops per mile and higher travel speeds, while buses in dense urban areas, like the 90s, tend to have more bus stops and travel slower.

One significant outlier is the 6 stops per mile, 20 miles per hour point you can see on the graph. That route is the J13, which only travels once per week at 6am on Saturdays, and flies through its route with relatively low congestion and ridership.

I'll be working on making graphs comparing WMATA to other transit agencies that publish data, as well as showing the different data for peak/non-peak and weekday vs. weekend.

Michael Perkins serves on the Arlington County Transportation Commission, though the views expressed here are his own. He lives in Arlington with his wife and two children. 


Add a comment »

About time they look at the issue!

High floor buses + a lot of stops = a lot of dwell time = limited attractivity compared to other modes!

by Vincent on May 20, 2010 1:32 pm • linkreport

Why not compare one of these variables to ridership?

by Tim on May 20, 2010 1:33 pm • linkreport

Also, it's been a while since I took statistics, but does this really qualify as a correlation?

by Tim on May 20, 2010 1:34 pm • linkreport

@Michael Perkins, what are you trying to show with this chart?

by Miriam on May 20, 2010 1:45 pm • linkreport

There's a correlation but it says nothing about causation or whether there's confounding variable.

And for that matter whether it's any more significant than noting that the average speed of cars on surface streets with stoplights is lower than the average speed on highways.

by ah on May 20, 2010 1:48 pm • linkreport

@Miriam, he's trying to show with data what a lot of us think (and common sense would expect) anecdotally...that as the number of bus stops increases, average bus speeds decrease. So if WMATA (or the other providers) want to improve bus travel times, one of the items (besides signal preemption or queue jumping lanes) to look at is reducing the number of stops.

@Michael, it'd be nice if you could label or somehow annotate select bus lines on the graph for comparison. For one, I'd be interested in how REX compares.

by Froggie on May 20, 2010 1:53 pm • linkreport

Second what ah said, isn't it more likely that more stops exist on slower streets (i.e. residential streets)?

Also, even if causation could be shown, the way I read this graph is that little speed is gained until you start spacing out spots a good quarter mile apart. I'm not sure that's politically possible or desireable.

Finally, I'll point out that of course you're using scheduled times/mph. Actual numbers might be well off from this.

by Reid on May 20, 2010 1:55 pm • linkreport


I think spacing stops 1/4 is both possible and desirable - but that may not happen at the expense of the existing stops. Just look at WMATA's Extra services (79, S9) which have that type of stop spacing and do indeed offer a) speedier and b) more reliable service, and in turn have seen c) increased ridership.

by Alex B. on May 20, 2010 2:10 pm • linkreport

This graph is a beautiful example of what can be created when technically inclined transit enthusiasts are given access to data in an easily readable form. Have you ever seen such an informative and data-rich graph come from WMATA itself? I haven't. In fact, I'd argue that it appears as if WMATA studies Edward Tufte's work just to do the opposite of whatever he recommends. But they've always been the ones with the data, hoarded. To get the data for this graph before bus schedule data were released in GTFS format, you'd have had to either hack at the PDF schedules or scrape the online trip planner, both of which would be rather onerous.

by thm on May 20, 2010 2:19 pm • linkreport

Hmmmm. More cold hard data. Very good.

@ the whiners: I'd like to point out that it is not trivial to produce a graph as this. We should all appreciate Matt's effort.

@ WMATA: Can you guys put up a server where all this kind of data can be collected?

@ ah: There's a correlation but it says nothing about causation or whether there's confounding variable.

I am not sure Matt is trying to show causation. I think he is trying to show cold hard data. We now know, from cold, hard data, that lines with more stops are slower. Nobody can argue anymore that that is just anecdotal evidence.

I am interested in seeing how travel speed is related to the speed limit on the route roads. My gut feeling is that there is little if any relation. Traffic lights and stops must have a larger influence.

by Jasper on May 20, 2010 2:44 pm • linkreport

I don't disagree that this is a worthwhile exercise. But let's make sure we're balancing the right things. Say you cut a bus stop and it causes half the riders to walk an extra eighth of a mile (which would be the minimum extra you'd walk) and with that you gain 5-6 miles in hour average speed. The walkers are adding about 2.5 minutes of walking to their commute (assuming an average of 3 miles an hour walking speed). Your trip would have to be 27 miles long for that 5.5 mph speed gain to make up for the 2.5 minutes extra it took to get to your stop.

Obviously personal travel time is way more affected by how long you have to wait for the bus in the first place, but that's just the point. It would be better for everyone if there were more stops but with a more reliable schedule and real time tracking system. Knowing I can leave a few minutes later and still catch the same bus accomplishes the same travel time savings as having the bus go faster.

by Reid on May 20, 2010 2:45 pm • linkreport

The author of this post and the perpetrator of this analysis is Michael Perkins.

Good work, by the way, Michael.

by Matt Johnson on May 20, 2010 2:46 pm • linkreport

We should all appreciate Matt'sMichael's effort. Sorry.

by Jasper on May 20, 2010 2:46 pm • linkreport

I have to admit this data isn't throwing out anything very interesting. More crowded lines go slower. Aren't they supposed to? That doesn't take away the value of an occasional express line that has limited stops.

A quarter mile is a long way to walk in the heat, and you have to factor the occasional hill as well. And not every bus makes every scheduled stop. There is some play in the system in there by design.

I've ridden this bugbear before, but what is amazing is more buses don't go from DC to Arlington and use the Roosevelt bridge. Just the 3Y and 16? 5A as well. You could move a lot of traffic from downtown that way and free up some space on the orange line. For example: an express bus from Federal Triangle to Rosslyn.

Another possibility is alerting drivers of the need to pick up passengers -- say a big blinking light and the stop. Sure, teenagers will set it off, but giving a driver 30 second or more to plan ahead could make a difference on speed.

by charlie on May 20, 2010 2:53 pm • linkreport

It's also interesting to note that the variance of speed appears to be greater for fewer stops-per-mile lines (we'd need more information to confirm this). This is important to note because it might indicate that simply lowering the number of stops for any given line won't necessarily increase its speed.

I think this chart is a good start to get people thinking, but I think a lot more variables need to be taken into consideration before we can really start spotting ways to increase efficiency.

by jon on May 20, 2010 3:04 pm • linkreport

I guess what I'm saying is that a lot of the frustration about the number of spots is not proportionate to the extra time they may (or may not) be actually adding to your commute.

It's worthwhile knowing how long an average trip is on each route. What is the speed limit? What about lights?

A lot goes into how long a ride takes, and any one factor may not achieve enough of a change to make it worthwhile to enough people.

I only say all this because I get the sense from most people who want fewer stops that they want fewer stops as a matter of principal. I think the analysis would have to be stop by stop line by line with clear objectives of overall average reductions in trip times not merely what would be less frustrating for some riders.

by Reid on May 20, 2010 3:11 pm • linkreport

Common guys, I work in transportation and have done such analysis in the past in what is one of the most advanced transportation markets in the world in the world

This is the kind of analysis I used to do. The only difference is the presentation which yeah used to be better because I and my company was paid to do the work.

Now if you want more explanation about the factors that influence bus speeds, then let's do an econometric analysis. One of you guys want to have a shot at it?

Kudos to Michael for everything he writes/analysis! he does an excellent job!

by Vincent on May 20, 2010 3:17 pm • linkreport

A bit of a tangent, but why is there a bus route that runs once a week early Sunday morning? Also, which bus route makes 10 stops per mile?

by Steven Yates on May 20, 2010 3:18 pm • linkreport

I certainly appreciate enormously Michael Perkins's efforts to gather and present real actual data. And it is good to know that the data support what we would intuitively believe to be true, namely that WMATA expects buses with more expected stops to travel more slowly.

My question is, now what? What should WMATA do with this information? Is it bus speed that WMATA should try to maximize, or ridership? To maximize bus speed, of course, all WMATA needs to do is eliminate all of the stops between the bus's origin and its final destination. But, in that case, how many people would take the bus? To maximize ridership, on the other hand, WMATA needs to balance accessibility (all things being equal, more stops mean more riders) and speed (all things being equal, taking longer to get there means fewer riders). How far are people willing to walk to a bus stop? How much time are people willing to spend on a bus trip from here to there? Those are the data we need next...

by Miriam on May 20, 2010 3:40 pm • linkreport

@everyone, thanks for all the comments. The graph was built using WMATA's GTFS data, read into a SQL database using a set of perl scripts I wrote. Other than writing the programs, this is all automated.

The answer I got back from the bus planners on why there are one-time-per-week routes is that they give people a ride to downtown before the Metrorail opens on the weekend.

10 stops per mile == the 90s, if I remember correctly.

I'm working on a much better graph using processing that will allow you to mouseover a datapoint to identify it. Also the dots will be sized based on trip frequency, and will be shaded based on peak vs non-peak.

And yes, I took the Tufte course.

I don't have a good database for ridership otherwise I'd include that in a different set of graphs.

The outliers on the "not many stops, still kinda slow" end are the DC circulators.

by Michael Perkins on May 20, 2010 3:44 pm • linkreport

Some highly-ridden buses spend a lot of time at stops because many people exiting the bus insist on using the front door instead of the back door. The back door should be fully utilized so that while people are getting off, others are embarking. Could raise the average speed a bit and also help buses run on time.

by Kouroush on May 20, 2010 3:45 pm • linkreport

I think in a lot of cases to say a bus route as an average speed and and average number of stops per mile is not the entire issue. The length of some of the routes is what makes them inefficient and always off schedule, and that's to me the more important thing to fix in the system. It's that they travel through different density areas with different traffic issues. That may be beyond this analysis to analyze the effects of breaking routes and adjusting routes, as was done with the 30's between Friendship Heights and Naylor etc.

I ride the 2B/C/G quite often, but only from the Ballston terminus a mile or so west. The problem is those buses are enroute from way out in Fairfax, and how anybody can expect a bus to be on time during rush hour over that route is beyond me. That is not directly affected by stops/mile but more the overall thought process of the routing.

by Lou on May 20, 2010 4:20 pm • linkreport

I'm a little late, but I couldn't resist working out the little math problem Reid screwed up above.

First, the relative benefits depend on the original speed, so let's say:

original speed = 20 mph
new speed = 25 mph
extra walking time = .05 h (3 minutes)

The time savings if your trip length is D is:
S = D/20 - D/25 = 0.01 D

So to save at least .05 hours we need 0.01 D > .05, or D > 5 miles.

If instead of 20->25mph we went from 10->15mph, we would only need a trip distance of 1.5 miles to come out ahead.

If you include time spent waiting at the stop, the case for removing stops becomes even stronger, since a 20% speed increase can allow 20% lower headways, so if your buses end up completely off schedule like mine usually are your waiting time also drops by 20% on average. Someone else can update the equation to include that.

Finally, everyone gets the time savings, not just the half who have to walk extra, so the total time for all passengers will improve (in the 20->25mph case) if the average ride is only > 2.5 miles. And extra finally, strategically removing bus stops will force much less than half the ridership to walk extra, so that might get us down to 2 miles or below for the threshold.

by J11 on May 20, 2010 5:38 pm • linkreport

Why not add express boxes to all routes that run less than 30 minutes between buses.

Just throwing out bus routes here


why not take away one trip and make it an express bus; I'm not looking at bus schedules so these are just examples

lets say a bus comes 7:05 and the next 7:20 then another at 7:35 why not make the 7:20 bus an express leaving the others alone and doing this to all routes.

You would be using the same amount of buses but speeding up the times of some (the ones which act as expresses)

by kk on May 20, 2010 6:36 pm • linkreport

Accidentally pressed the button before done

You could have the "express bus" displayed as another route
X22, 900/922, 38B2, L11/L22/L44 or whatever can be thought of and just run them at specific times of the day.

You could add it to every single bus that does not run like 30, 45 or 60 minutes and be done with it.

by kk on May 20, 2010 6:41 pm • linkreport

One thing that would be interesting is an apples-to-apples comparison of express and local buses on the same segments at the same time of day. This would provide some, albeit limited, information about the effects of reducing stops on speeds.

by Chuck Coleman on May 20, 2010 7:58 pm • linkreport

Michael, you might be interested in Tableau Public software. It would save you some coding time, and let people answer some of their own questions.

by biggerbox on May 21, 2010 11:16 am • linkreport

Michael, can you publish the regression output? Or possibly email it to me? Good work!

by Rob Pitingolo on May 21, 2010 12:31 pm • linkreport

@Rob: I can send the text file containing the output. There isn't a regression because I didn't calculate one. There are software packages that will do it.

by Michael Perkins on May 21, 2010 12:36 pm • linkreport

Never, ever get in front of an express bus!

But there's a reason for many stops: Some people can't walk far. ADA requires routes to have minimum walking distance.

Does DC Metro provide special busses for accessibility?

by Crissa on May 21, 2010 5:34 pm • linkreport

@crissa: There's a need for a lower walking distance to get to a stop, but a route that has 8 stops per mile on average will have 10-12 per mile for a portion. That's only 500 feet, shorter than the length of a Metrorail platform. The ADA requires paratransit to be provided within 3/4 miles of a stop, so that's a good estimate of a maximum distance the law thinks people should be able to go to get to a stop.

Not that 3/4 of a mile is a reasonable distance for the spacing of all stops. That's way too far. But stopping every 1/10 of a mile is too often.

by Michael Perkins on May 21, 2010 5:48 pm • linkreport


You totally made my day. I've been searching for a software like Tableau for the past 3 weeks and then, boom, here it is. I was researching "R" and then saw Michael Perkins reference Processing and downloaded it. But this takes the cake. Brilliant software.

by michael on May 24, 2010 5:09 pm • linkreport

What about spreading buses out throughout all streets.

Take central DC and buses going North/South


North Capitol, New Jersey Ave, 5th, 7th, 9th, 11th, 13, 14, 16th, Conn, New York, Rhode Island, Florida Ave.


North Capitol, 8th, 15th, Trinidad, Montello, BladensburgRD, 18th, 19th, Mass, Potomac, Penn, Minnesota.

Why not spread routes out instead of them all going along the same streets and have them all travel every 2 streets apart that way more people are closer to a bus stop.

Some parts of the city you can go to every cross street and catch a bus but in other you will make walking for miles to find another street with buses.

by kk on May 24, 2010 5:46 pm • linkreport

Add a Comment

Name: (will be displayed on the comments page)

Email: (must be your real address, but will be kept private)

URL: (optional, will be displayed)

You can use some HTML, like <blockquote>quoting another comment</blockquote>, <i>italics</i>, and <a href="http://url_here">hyperlinks</a>. More here.

Your comment:

By submitting a comment, you agree to abide by our comment policy.
Notify me of followup comments via email. (You can also subscribe without commenting.)
Save my name and email address on this computer so I don't have to enter it next time, and so I don't have to answer the anti-spam map challenge question in the future.


Support Us