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


Where the Metro riders are, and aren't

How full are Metro's trains at any point in the system? What routes do riders take when confronted with a choice between two transfers, or between a longer one-seat ride and a transfer? Last year, in discussing maps I created about the proposed Blue Line reroute, a reader asked about this, but Metro hasn't collected the data.

To answer this question, classmate and frequent GGW commenter Reza and I created a non-scientific survey, and used the data to build a diagram of passenger traffic on Metro system. This survey was unscientific and the results should not be considered as absolute facts. It does show potential trends, but a larger and broader sample would be necessary to validate these results.

A diagram of estimated ridership on the Metro system. Click to enlarge.

The survey first asked respondents to state a certain preference, like the top factor they use to decide on a route. Later, it presented a specific scenario and asked respondents which route they would choose. Overwhelmingly, survey respondents primarily prioritized getting to their destination in the shortest time. A plurality of 44 percent chose minimizing transfers as the second most important factor. For the third choice, a plurality of 37% said, given the option, they'd choose a line with more frequent trains.

Based on the responses, Reza and I created a decision tree to assign the trips from the 2007 Ridership Survey to the links of the Metro system. On the segments of the system without alternative routes, the diagram is 100% accurate from Metro's origin and destination data. For example, ridership between Takoma and Silver Spring is definitely almost twice as high as between Silver Spring and Forest Glen, which is why Metro turns back half of Red Line trains at Silver Spring. Inside the area with alternative routes (bounded by Fort Totten, Metro Center, L'Enfant Plaza, Rosslyn, and Pentagon), the diagram relies on the non-scientific and possibly non-representative survey.

A geographic representation of estimated Metro ridership. Click to enlarge.

This information is very important to helping Metro make good decisions about service. Dropping the percentage of rush hour Blue Line trains at Rosslyn 40 percent to only 20 percent, as Metro proposes, would make the most sense if about 80 percent of riders at Rosslyn were on the Orange Line. But some readers of Greater Greater Washington and elsewhere commented that their Blue Line trains between Rosslyn and Arlington Cemetery seem pretty crowded, and that they probably would become even more so if Metro halved the number of trains.

Metro's May 2007 ridership survey lists the number of trips made from any station to any other station, on average, but not which path riders take to get there, when they have a choice. We don't know what emphasis riders place on factors transferring versus travel time, and therefore can't ascertain what routes people would choose if given an option. Of course, a trip from Shady Grove to Dupont Circle can only happen via the Red Line, but from Van Dorn Street to L'Enfant Plaza, the rider has to choose between the one-seat Blue Line ride and a transfer to Yellow, or a trip from Woodley Park to Prince George's Plaza involves a transfer at either Fort Totten or Gallery Place.

Metro needs accurate models to make decisions about service levels given its budget and infrastructure constraints. They periodically take a statistical sample of riders to determine where people are boarding, where they're exiting, and how they get to and from stations. But they have little data about how people get from point A to point B. One method that Metro uses to determine the ridership on certain line segments is to station workers on the platform to count passengers on trains, but this doesn't capture all information about route choices. WMATA should consider adding those questions to their ridership surveys.

The ridership per link estimated by our survey and assignment model show some interesting relationships.

As mentioned above, ridership drops by half on trains going northbound through Silver Spring. The decision to short turn trains there (because of the presence of a pocket track) was a good one. The phenomenon does not repeat itself on the other side of the Red Line. Volumes never drop significantly at any one stop, although they do taper as the line approaches Shady Grove. During rush hours, half of all trains turn back at Grosvenor, but unlike at Silver Spring, there is no major drop off in volume there. In fact, there are more riders in the link south of Grosvenor than there are in the link south of Silver Spring and ridership is higher at every single link north of Grosvenor than it is on the link between Silver Spring and Forest Glen.

Downtown, the Red Line is very busy. As one would expect, there is a significant jump in ridership at Union Station when coming from Glenmont. Ridership jumps by almost half from the link north of Union Station to the link south of Union Station. It might be worthwhile to find a way to insert a pocket track into the southern tip of Brentwood Yard and run some rush period trains from Shady Grove to New York Avenue.

In Virginia, there are significant drops in ridership west of Ballston and west of West Falls Church. West of Ballston, ridership drops by approximately one-quarter, and then by another third west of West Falls Church. Currently Metro does operate some trains from/to West Falls Church during peak periods. It might be helpful, especially after the Silver Line opens, to construct a pocket track in the median of Interstate 66 between Ballston and East Falls Church. This would allow some trains from or to downtown to serve the crowds of the Wilson Boulevard corridor. This will be especially important once the SIlver Line starts to reach ridership targets because Arlingtonians will find it harder to get on already crowded trains from the suburbs.

Volumes also drop on the Blue and Yellow Lines south of King Street. The combined ridership south of King Street is 20 percent lower than on the segment north of King Street.

What about Rosslyn, the spot which started this whole endeavor? Based on our analysis, of the riders traveling on the two links immediately outbound from Rosslyn, 62 percent are on the Orange Line and 38% are on the Blue Line. This matches closely current service levels.

At Pentagon, a similar look shows us that of passengers traveling on the two links inbound of the station, some 54 percent are on the Blue Line and 46 percent take the Yellow Line Bridge. This also closely matches current service levels.

This brief analysis demonstrates some of the difficulties with understanding ridership patterns on the Metro. It shows why it is so important for Metro to find some way of surveying patrons on how they travel, not just where they travel. Unfortunately, because of the limitations on our surveying it is difficult, if not impossible, to draw substantive conclusions about ridership patterns themselves. However, it does offer an interesting glimpse at a better way to plan for service alterations.

Matt Johnson has lived in the Washington area since 2007. He has a Master's in Planning from the University of Maryland and a BS in Public Policy from Georgia Tech. He lives in Greenbelt. Heís a member of the American Institute of Certified Planners. He is a contract employee of the Montgomery County Department of Transportation. His views are his own and do not represent those of his employer. 


Add a comment »

Fascinating! Thanks, Matt!

by tom veil on Jul 28, 2009 10:29 am • linkreport

GREAT map! It shows how underutilized the non-TOD stations are on the Eastern side of the city/MD.

by SG on Jul 28, 2009 10:49 am • linkreport

Great work!

Even though it wasn't that long ago, my memory has failed me in that I can't recall how the survey was conducted. It might help just to touch upon that so that people taking this data any further would understand the foundations upon which it's built. Was GGW the primary tool for getting responses?

by Bossi on Jul 28, 2009 11:15 am • linkreport

If anyone wants the full report, they can contact me at

That includes methodology and such.

by Matt Johnson on Jul 28, 2009 11:19 am • linkreport

Since Metro collects data on riders both entering and exiting through its fare system, isn't the data all there?

by Anderkoo on Jul 28, 2009 11:34 am • linkreport

Really interesting analysis. Thanks!

by Adam F on Jul 28, 2009 11:36 am • linkreport

If there were a pocket track after Ballston, Metro could create a peak-time Arlington Local service running from Ballston to the Airport, or an A-A Local to King Street. It could be the only station to serve Arlington Cemetery whenever it's running. This would be great for Arlingtonians who don't like crowds; would there be any other merit?

by Beau on Jul 28, 2009 11:36 am • linkreport

I don't believe the existing data is tied to origin/destination to make specific pairings. So essentially you get volumes at stations, but don't know what other stations the traveler specifically originated from or is destined for; nor the path they'll take in between.

I don't think converting the existing system to collect O-D pairs would be difficult, but I'm pretty sure I read somewhere that Metro has explicitly decided not to... I think it was privacy issues. Others may be able to correct me on this. One caveat is that even if the fare system was converted to gather O-D pairs, its present configuration wouldn't capture the route people take.

by Bossi on Jul 28, 2009 11:40 am • linkreport


No it is not.

For instance, I used to commute from Prince George's Plaza to Union Station. My SmarTrip card would have told Metro this, but Metro would not have any way of knowing whether I transferred at Fort Totten and used Red via Brookland or transferred at Gallery Place and used Green via Columbia Heights.

Similarly, they don't know if a trip from Pentagon to Capitol South generates a single-seat ride on the Blue Line or a two seat ride on Yellow and Blue.

That was why we needed to determine rider preference and create a decision tree in order to assign riders to certain links.

The main reason I suggested a pocket track at Ballston is because of the ridership jumps. Once the Silver Line opens, trains coming from Ashburn/Ryan Road will have had plenty of time to fill.

According to riders on the A-Shady Grove Route, after the Metro crash with no Grosvenor short turns, it was impossible to get on a Red Line train at, say, Van Ness. They were already too full.

Ridership is only going to increase. What Metro does with its data is up to them, but using a study like this would go a long way to figuring out where good locations for pocket tracks are.

by Matt Johnson on Jul 28, 2009 11:41 am • linkreport

Points taken. Just thinking out loud here -- if Metro decided - perhaps temporarily - to collect O/D data, maybe the task of conducting a study could be boiled down to counting transfers at the five possible stations? Could strategically-positioned video cameras do the trick for later machine or human analysis?

by Anderkoo on Jul 28, 2009 11:54 am • linkreport

You'd have to look at how the stations are aligned & the paths people would take... but I'd think getting *everyone* would be a daunting task. Some people may get off the train and go the wrong direction; there could be multiple routes people may take; routes may overlap with those entering/exiting the station; and simply watching all those faces could be tricky.

However... by numbers alone: with a statistically significant sampling I think that could work. It might be done manually with surveyers simply asking where the travler what stations they started at & are going to. It'd take a 30-second converation, of which half of it is spent on the surveyor identifying him/herself. The tough part would be getting travelers to stop for those 30 seconds as they briskly travel between platforms. When I'm transfering, my mentality is always "there's a train waiting for me right now, and if I don't make it I'll miss it".

Personally, I think Matt's approach is just as well... posit to people, while they're sitting at the computers, to think about how'd they'd act. It may not be as accurate as catching people while they're actually doing it, but it'd be far easier to conduct & organise.

Or another option would be instead of on-site interviews, hand out business card sized forms with 4 lines to fill in: origin, transfer 1, transfer 2, destination; and then put a box at the exit gates. The back of the card could explain the reasoning for the data & a phone number or mailing address.

by Bossi on Jul 28, 2009 12:07 pm • linkreport

What's interesting to me is the strength of the ridership in the "favored quadrant" and the weakness of the suburban legs elsewhere.

by Paul on Jul 28, 2009 12:27 pm • linkreport

as anyone who has lived in DC all thru the history of Metro can tell you, Ward #9 citizens prefer driving cars over transit. Just look at the lack of transit oriented development around their stations- they are isolated and not at all comforting places- and in this way they actually discourage ridership. If they would only see /wake up to the idea that density around transit hubs increses safety , things would go for the better.

I also place blame on the metro board members- who all drive and almost NEVER use Metro.

by w on Jul 28, 2009 12:31 pm • linkreport

Great map. Does Metro break the ridership data down between peak periods and off-peak periods? If so, it would be interesting to see the difference.

It's worth noting that if the number of people using every station were exactly equal, you'd expect about a 20% drop in ridership west of Ballston and a 33% drop west of West Fall Church: of every 100 people approaching Ballston from the east, 20 would get off at Ballston, 20 would get off at East Falls Church, and so on (and only a few people would get on at these stations and go west). The drop in ridership as you approach the ends of the lines is not evidence that the outer stations are underutilized. The number of riders between Vienna and Dunn Loring will always be tiny compared to the number between Farragut West and McPherson Square.

by Johanna on Jul 28, 2009 1:04 pm • linkreport

I also notice that there's a pretty big jump percentage-wise on the Green Line at Suitland -- I'm assuming that people are commuting to that station for jobs in the Federal Center. I would have assumed that commuters from Southern Maryland that drive to Branch Ave would make the Green Line taper more as it went away from the city as opposed to a big jump.

by Reza on Jul 28, 2009 1:19 pm • linkreport

This is great analysis by Matt. However, I take objection ot the following statement:

"Dropping the percentage of rush hour Blue Line trains at Rosslyn 40 percent to only 20 percent, as Metro proposes, would make the most sense if about 80 percent of riders at Rosslyn were on the Orange Line."

This statement assumes the decision to build the Silver Line wasn't already made. Since that decision has already been made, slots need to be freed up for the Silver Line and then the question becomes how do you go about doing this the most efficient way. With that in mind, the Blue line diversion during rush hours is the best solution given the constraints. The system through the core is so heavily utilized, there are no win-win solutions. Sombebody has to lose and in this case its the people going to Rossyln, Foggy Bottom, Farrugut West and McPhereson Square. No amount of complaining from those who lose is going to change the reality that somebody has to lose.

by Dharm on Jul 28, 2009 1:57 pm • linkreport

This looks like a good TRB paper... abstracts due Aug. 1!

by J on Jul 28, 2009 2:15 pm • linkreport

While I'm sure this would set off some conspiracy types (and I'm not sure I'm comfortable with it either), it'd be fascinating if WMATA actually collected station to station data. You could say 'Of the X thousand people who got on at Ballston today, Y got off at Foggy Bottom, Z got off at Metro Center' and so on.

This could potentially identify good locations for new routings and service increases. Track it even more closely and you could know which exits from which stations are busiest, to know where to add fare gates (Capitol South seems to jam up more than it should).

This data already exists, since it's on the fare cards to charge correctly. I assume that the fare gates have no way of reporting the data, though, and it'd be prohibitive to upgrade them. It'd be nice, though.

by Distantantennas on Jul 28, 2009 2:28 pm • linkreport

@Anderkoo, @Bossi,
We don't need to use video cameras.

There are more than 6 stations anyway. For instance, someone coming from Franconia and going to Mount Vernon Square could transfer at King Street or Pentagon (both marked on the map with a transfer target) or they could transfer anywhere in between or at L'Enfant.

The point-to-point data is already collected. That's how I knew how thick to make the lines. Every so often, Metro conducts a survey of a statistical sample of riders. They're asked several questions, including:
(a)The address of their origin for THIS trip,
(b)The mode they used to get to the station where they entered the system (walk, bike, bus, drive, rode with a driver, etc),
(c)If they used a bus, what was the number and operator,
(d)The station they will exit the system at,
(e)What mode they will use to get from the station to their final destination,
(f)If a bus, what number and operator,
(g)The address of their final destination,
(h)The purpose of their trip,

All WMATA needs to do is ask them to mention which station(s) the'y be using to transfer (if applicable), and we'll have real data on route choices. So, since WMATA already conducts a survey, all they need to do is rethink the questions in order to get better data.

Metro does collect data on time of entry. However, I do not have access to that data. My methodology would need to be reworked before that data could be included, however, because I would need to have two lines for each segment (representing direction of travel).

You make an excellent point. And I made exactly the same point some time ago:

That will be the case once the Silver Line opens. However, until that happens, it is not necessary for WMATA to reroute trains on the Blue Line.

That said, I'm not positing anything other than the fact that WMATA needs to try and collect data about how people travel (route choice).

I argued before, and I'll continue to argue today, that the time loss for travelers going to the Rosslyn/Farragut Areas is not huge. A simple readjustment of time of arrival at Franconia or Van Dorn will nullify any loss of time (increase in travel or wait times).

However, if this estimation is accurate, every Blue Line train going through Rosslyn is going to double its load. That will increase dwell times between Rosslyn and L'Enfant and will just make riders more irritable.

On another note, WMATA does keep track of metrics for level of service. One of those is train crowdedness. They try and keep train loads under a certain threshold. If the Blue Line reroute puts the Blue Line over that threshold (and I say if, I don't know that it will), it will start failing on LOS measures, and they'll have to take steps to rectify that.

That data already is collected by WMATA. I could tell you how many people *on average* travel from any station 'x' to any station 'y'.

What I couldn't tell you (prior to this survey) was which route people took to get from x to y.

by Matt Johnson on Jul 28, 2009 3:40 pm • linkreport

Really? Interesting. I've only ever seen data get published with overall boardings (So many hundred thousand today) or with individual station boardings in a day. It's good they have more detailed data - just never seen it published. Thanks for the info.

by Distantantennas on Jul 28, 2009 5:37 pm • linkreport

A little history: Original plans (system map displayed aboard trains when first segment opened) showed the future Yellow line operating between points west of Van Dorn Street and Greenbelt. The future Blue line was shown operating between Huntington and Addison Road.

When WMATA was ready to extend service south of National Airport in 1983, WMATA did not have enough rolling stock to extend the Blue line to Huntington. After analyzing the rolling stock distribution they discover they could open the segment if they ran the trains as Yellow line trains.

As the 2 and 3k series cars became available and WMATA opened Van Dorn Street, WMATA chose not to switch the line colors back to the pre 1983 line colors.

The 2 and 3k series cars became available well after service was extended to Huntington.

As to the issue about a pocket track north of Union Station. The original plans for the New York Avenue station called for converting the now abandoned northbound track B1 into a passing track. Plans called for using it to short turn trains at Union Station. The track was not converted to a passing track because of funding. WMATA can in the future when money becomes available connect the abandoned track to the existing mainline.

by Sand Box John on Jul 28, 2009 6:24 pm • linkreport

This is great! I'm glad I participated in the survey.

by Lindemann on Jul 28, 2009 6:25 pm • linkreport

Great work, a couple of follow-on thoughts:

1- Could you visualize how "inboundy" or "outboundy" each station is? I.e., what percent of boardings are headed towards downtown, vs. away from downtown?

2- Is there a way to visualize how much turnover there is at each station? This would be a useful supplement to loading. For example, it looks like loads north and south of U Street are fairly similar, which indicates that boardings and deboardings are roughly equal - but how many boardings and deboardings were there? How high was the turnover?

I guess in other words, is there a good way to visualize the connection between ridership by link (well done here) and ridership by node/station?

by jantos on Jul 29, 2009 9:23 am • linkreport


That is an interesting question. Certainly the downtown transfer stations on the Red Line have very high turnover as people finish the extent of their suburb-to-downtown commute on the line and either exit the station or transfer to another line to complete their journey.

It seems likely that inner stations that have a lot of TOD nearby (Columbia Heights, R-B corridor, Bethesda) will have a more equitable distribution of people using it as either an origin or destination station compared to single-use stations (commuter stations are primarily used as origin stations while Federal City stations are primarily used as destination stations during the morning commute).

Another thing to consider is that typically the last few stations on each line that are TOD-oriented (and also usually lack parking) will mark the limit to the extent of people traveling outbound to jobs or other destinations around those stations. Case in point, King Street on the Yellow/Blue lines, Bethesda and Medical Center on the Red Line A route, Silver Spring on the Red Line B route, Ballston on Orange Line, PG Plaza (though this station does have parking) on the Green Line.

I'll defer to Matt on whether this model can be used to more definitively answer your question, however.

by Reza on Jul 29, 2009 12:29 pm • linkreport

I agree with some of the above comments: directionality would be a great thing to add to this map. I would love to see what percentage of the trips between stations are in which direction. This sort of analysis is very important for maximizing the current system's capacity. Lots of empty trains are going east in the morning due to the lack of Metro-accessible job sites in Prince George's county. It would be great if this map could help illustrate where WMATA's core capacity is really reaching its limit and where there are opportunities for moving more people on the existing system.

by Michael on Jul 29, 2009 3:18 pm • linkreport

It looks like there ought to be a downtown relief route. Perhaps just grabbing the local short-hop traffic, or diverting people off the Metro further from downtown....

Exclusive lanes for streetcars anyone?

by Nathanael on Jul 30, 2009 4:18 pm • linkreport

This looks like a very good case for a Mount Vernon Sq to Rosslyn Link, rather than the conventional Stadium-Amory to Rosslyn separate Blue Line.

by James D on Jul 31, 2009 11:00 am • linkreport

James D,

Not sure I follow your logic. For one, the data shows that extending the New Blue line to Union Station would be extremely beneficial as well. Furthermore, the idea of simply having a stub line terminate in the middle of DC without connecting to anything else would be an operational nightmare - that's why all the proposals continue on to Stadium-Armory (or thereabouts).

by Alex B. on Jul 31, 2009 11:53 am • linkreport


1. With better data, we could indeed visualize how "inboundy" a station is. However, with the data provided by WMATA, we don't have enough information at this time.

2. Now, this would be easier with better data from WMATA, but is theoretically possible with the data already in my possession. However, my model was not designed to spit out those numbers. It would require a redesign of the model to get that, and that would take a bit of time.

Anyway, should WMATA provide better data, look forward to more interesting posts in the future.

by Matt Johson on Aug 5, 2009 12:03 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