The MARC Train System in Graphs

Recently I've started using the MARC train system a lot (precipitated by wrecking my car and needing to get to work). It's more expensive than driving and my commute is not bad (I have a reverse commute), but I've found it's really a pleasant way to get to and from work. I've also found that there are WAY more people riding the train that I'd thought. I stick to the Baltimore area, but I'm super impressed about how many people are using the trains every day.

The MTA does not publish ridership statistics on the MARC trains, so I emailed their “Contact Us” page. Much to my (pleasant) surprise, they replied almost immediately and were more than happy to share information. In a world where most places ignore your emails or take days to give you a non-answer, the MTA is a great exception. Anyway, I learned some interesting facts:

1. MARC train get a lot of riders
2. Generally, more people are riding the MARC now than ever before
3. It's impossible to make strong statements about how much ridership the MARC is getting

Statements #1 and #2 come from the data that the MTA supplied me. Statement #3 comes from their ridership counting methodology. Here is exactly how they determine daily ridership: on the first Wednesday of the month conductors count everyone who gets on and off at each station. Why Wednesday? Because Wednesday is typically the peak usage day for the system. So, as you'd expect, the data is SUPER noisy and absolute numbers are unreliable. However, after looking it over I'd say that it's probably pretty good from a general trends standpoint. Getting better data is probably possible, but very expensive and of questionable usefulness. I'm pretty sure that the data the MTA does get is good enough to manage the system.

On the other hand, what the MARC data lacks in detail on a day-to-day basis is made up for in it's historical value. The MTA sent me a spreadsheet with their data for monthly ridership counts going back to October 2001! They included their 12 month moving average data going back to February 2002! And, in the spirit of sharing, I've put that spreadsheet up for anyone to download.

It's a lot of data, and whenever I'm confronted with a lot of data I try to graph it to make some sort of sense of it all. So since they sent me this data I've been manipulating it in different ways, trying to see what I can learn and how best to describe the whole system.

(Note: The following discussion assumes that you have a passing knowledge of the MARC train system. Wikipedia and the MTA have plenty of information, but basically you should know that there are 3 passenger train lines – the Penn Line, the Camden Line, and the Brunswick Line. The Penn Line operates on the Amtrak-owned Northeast Corridor tracks, the Camden Line and Brunswick Line are at the mercy of CSX freight trains. All lines terminate at Washington Union Station. The Camden Line terminates at Camden Yards in Baltimore. The Penn Line passes through Baltimore Penn Station and continues towards Aberdeen and terminates at Perryville. The Brunswick Line has two spurs, one to Martinsburg, WV and the other to Frederick.)

A final note on this data. The MTA counts total station usage (i.e. boardings plus alightings), so the number of passengers is the total usage divided by 2. All of these numbers are daily numbers, but counted on peak usage days, so the average is discounted by some amount (which I won't even guess at). The MARC trains operate weekdays (an average of 250 days per year).

First off, total station usage by station. I have split Washington Union Station into 3 pieces, one for each of the rail lines.

First of all, you're probably thinking “Wow, I never knew that there were so many MARC stations!” and “Wow, everyone's going to DC and they're doing it on the Penn Line.” The Penn Line blows the Camden and Brunswick lines out of the water in terms of total ridership. In June 2009, the Penn Line accounted for 63.5% of total MARC system usage. There are many reasons for this, including facts like better on-time performance because the Penn Line is on Amtrak-owned tracks, good station locations in population centers, and the whole idea of the rich getting richer – the Penn Line is highly used so more capital investment is made on the Penn Line to length platforms, build parkings spaces, etc.

What does total system usage look like? Well, in June 2009 the 12 month moving average of total usage was running at 34,274, which indicates over 17,000 passengers on peak days. That's a lot of cars off the road when you think about. Especially considering that roads and traffic delays are fairly non-linear, the more cars on the road the worse the effect of each marginal car. Since most of these trains run for commuters to reach DC, traffic could be far worse without this service.

Just how many of these passengers are going to DC? From the data I have, I can't answer that question. I can, however, answer a related question, “How many of these passengers arrive or depart from Union Station?” And, “How has that changed over time?”

This chart shows some very interesting trends. Over the course of the last 7 years, the Brunswick Line has increased the proportion of trips that do not involve Union Station from 30% to 40%. That's very helpful for the keeping Union Station from being overwhelmed (spoiler alert: Union Station is pretty much overwhelmed by commuter traffic). I think it's pretty easy to explain this trend though. The Brunswick Line makes stops at Rockville, Kensington, and Silver Spring – good place to get off the MARC and get onto the Metro. Many of the commuters on the Brunswick Line may well be going to DC, they just have better options than Union Station.

The second interesting trends on that chart is that the Camden Line's fraction of travelers not going to Union Station peaked in 2008 and has since decreased. My estimation there would be that during high gas prices in 2008 a few more people started to use the Camden Line to commute to downtown Baltimore. Baltimore's traffic and parking situation is nothing like DC's, so when gas prices eased people reverted to their old ways. Honestly though, this could just be a strange blip in the data. The Penn Line has such noisy data for trips that do not involve Union Station that it's hard to understand. When you look at total trips (instead of a fraction of the whole), the data is noisy enough to make you think there is just something wrong with it.

There are some strange dips and peaks in the Penn Line data while the Camden Line and Brunswick Line show much smoother trends. What is good news for the MARC system is the combination of these two graphs. Trips that do not involve Union Station are increasing faster than overall system growth (as evidenced by the fact that total number of trips has increased as well as the fraction of the total). This means that instead of a system that runs trains full of DC commuters one-way and empty trains the other, you are starting to get a few more paying customers on those trains that would otherwise be empty.

Now for a look at the total system usage over the last few years:

If you're a fan of public transit you've got to love a chart that looks like that. The R2 value is 0.96 for a line of increasing ridership from 2002 through 2009. The fit line says every day from February 2002 to June 2009, an additional 1.75 passengers started riding MARC trains. That's great news for the system, but it's not unconditionally good news. I've had a hard time finding official numbers on what “capacity” actually is for the system, but the numbers I have seen are striking similar to actual usage. That means that system is actually operating near capacity and could probably increase ridership further with additional resources.

This is a lot of growth, but where is it coming from? Are there stations that have actually seen fewer riders? In general, growth is everywhere; in specific, the Penn Line. Over the past 5 years the Penn Line has added more usage than the Camden Line has total.

Line	12 Mon. Moving Avg. Ridership		Increase	Increase (%)	Average Annual Growth Rate
	June 2004	June 2009
Penn	16,273	21,762	5,489	33.73%	6.0%
Camden	4,255	4,768	513	12.05%	2.3%
Brunswick	6,765	7,744	979	14.47%	2.7%
Total	27,293	34,274	6,980	25.58%	4.7%

Here are a few charts that show a “then and now” look at the three lines over the last five years. I've maintained the same scale between charts to illustrate how the Penn Line dominates the MARC system. I do not include Union Station simply because it would ruin the scale of the chart.




And finally, just because I have the data and can plot it, here is a look at Penn Line station usage over the last 7 years and Union Station usage by the three MARC lines.




For the Penn Line data, the most impressive part of this data is that even after the financial crisis and the tremendous drop in gas prices, the Penn Line has not lost passengers. Perhaps these passengers are like me, they started riding the train by necessity and they kept riding it because it's just a lot less stressful than fighting traffic. Honestly, I think the number one thing constraining ridership at some of these stations is just parking availability. The plans to build additional parking at West Baltimore will likely increase the ridership there.

Meanwhile, the data on Union Station shows just why that train station is able to support so many businesses and shops while Baltimore's Penn Station can barely support a snack stand and a souvenir store. Actually, that is a bit of a unfair comparison, Penn Station seems to have more than enough traffic to support a few more commercial enterprises – but I'll have more thoughts on that in another blog post.