Trying to adjust the Census' Commute Mode Share calculation for DC

The Census Bureau's recent report on the growth of bike commuting over the 2000-2010 period got me thinking about the criticism of the data, of which there are basically two (not counting sample size):

1. The ACS only counts the “primary” mode. It does not count commuters as bicyclists if they rode only part of the week.

2.  The ACS does not count commuters as bicyclists if they rode their bicycle to/from transit and the transit portion was longer.

I don't see the first criticism as much of an issue. If there is reason to believe that there a signficantly more 1 and 2 day bike commuters than there are 3 and 4 day bike commuters (like me), I haven't heard it. In other words, it should all be worked out in the averaging and I don't believe it distorts the results.

But the 2nd criticism is pretty fair, and Brian McKenzie of the Census Bureau admits that during this recent interview, which I only partially listened to. 

So is there a way to capture that? Not really, but one piece of data from the 2012 Metrorail Passenger Survey is that 1% of Metrorail users are arriving by bicycle might help us estimate it. aAcouple of caveats first, that data is not availble by jurisdiction the way it is for walking, but since it's unclear to me if it is underreporting or overreporting the number for DC. Another flaw is that it only studies how people get to Metro, not how they leave. 

The number of daily DC metrorail users, according to the ACS, was 70,781. Of which 1% would be 708. Adding that to DC's current estimated number of bike commuters would bump DC up to 4.4% from 4.1%. That's a floor. 

Of course, that doesn't count people who bike to Metrobus, Amtrak, MARC or VRE. [I'm going to assume that very few (~0) people use their bike in combintation with driving, carpooling, taxis, motorcycles or walking]. But if we assume that a similar 1% of all transit users access the system by bike, then the percentage of DC commuters who use a bike for all or part of their commute goes up to 4.5%. And, that's still undereporting, because it only counts people who have a bike-transit-anything commute, but not people who have an (anything but bike)-transit-bike commute. 

I'll be the first to admit that this is not a particularly robust calculation and that much better data would be needed to really tease out the percentage. Furthermore, it does involve allowing for someone to be simultaneosly a bike commuter and a transit commuter, (when I used to do a combo commute, that's how I thought of myself) which means that the total of all reported groups will be more than 100%. but I think the idea that the Census Bureau's measurement is undercounting bike commuters, by ignoring combo-commuters, by 10% is not too far off.

BTW, using the same methodology walk commuters are being undercounted by much more as 32% of all Metrorail riders are arriving on foot (and again that is metro-wide. I'd expect that number to be much higher in the city where there are fewer parking garages). If we add them in, assuming the same breakdown across all transit users, the percentage of DC residents who walk part or all of the way to work goes up from 11.9% to 24.3%.

And it's only fair to mention that this means that car commuters are likely being undercounted too. I'm not sure how many DC residents drive to transit, but there must be some, especially when considering "drop offs". 

Comments

[Purple Eagle]

I met a guy once who drives from somewhere in suburban MD to College Park and then bikes on the Anacostia Trib trail to DC. But he's presumably not a DC resident. There is (car) parking at Ft. Totten and Rhode Island metro stations and a few others in DC so there are probably some DC residents to park there.

[Matt]

My problem with bike commuting census data from the ACS is the estimated margin of error. I was playing with some of that recently (prior to this report though) and in some cases the +/- margin was greater than the total count. So you'd end up with a census block that had 175 bike commuters, +/- 300. Not very helpful!

[ACyclistInTheSuburbs]

On the one hand there are lots of newb bike commuters who only do it occasionally, or who do it only on Fridays (to avoid having to deal with non casual office attire, and to ride when traffic is lighter) but there are lots of regular bike commuters who take day off biking now and then because of extreme weather or to bring stuff to the office, etc, so your guess they offset is probably right. It would be a nice topic for someone to do a special study.

[7]

FWIW - I maintain a small incentive program in my office for bike commuters. According to my records - 10 of 32 bike commuters in my program self-declared that they commute only 1-2 days/week. Seven declared to commute 5 days/week. Only 2-3 do a bike/transit mixed commute. These are commutes to downtown Silver Spring.

[Matt]

@ ACyclistinTheSuburbs - Your suspicion (and that contained in the original post) are probably accurate if you're measuring trip counts.

However, if you're measuring "people who bike commute at least once a week" then we're underestimating the overall numbers. I think this second number is possibly more important than trip counts.

[SJE]

I think that focusing on mode by distance travelled or number of days used is similar, and similarly wrong. The way the data is collected drives the way in which it is used/misused.

The current way the data is collected leads to a systemic undercounting of the diverse ways we get around.....which affects spending and prioritization....which merely reinforces driving.

If I drive 3 days a week and ride 2, for whatever reason, the census would have me as a driver and move $ towards car infrastructure. But I still want a safe route to work when I ride.

[washcycle]

But if you drive 2 days a week and ride 3, the census would have you as a cyclist and move money toward bike infrastructure.

Perhaps a better measure of mode share would be to try to determine what percentage of your average weekly commute trip is done per mode.

So mine might by 75% bike and 25% drive solo. My wife might be 90% transit and 10% walk. Etc...

[black jack]

IIRC the Census question asks what was your primary mode of transportation to work "last week," as opposed to "in general." So it not only allows only one mode as a response, but takes a snapshot rather than an average.

I typically commute by bike, but if "last week" it was 20 degrees, or snowed, or poured rain for three of the weekdays, I would have to answer that my primary mode was transit, not biking.

But over the course of a year, transit probably accounts for only about 25% of my trips to work.

[DE]

Before I worked at a place with showers, when it was hot or raining I would drive to the Courthouse area, park for free, and ride my bike to Rosslyn (where parking was expensive). But yes, the number of people who do that type of thing is probably statistically insignificant.

[washcycle]

black jack, they do the survey all year, so for every person over-reporting their biking during good weather weeks there should be someone under-reporting it during bad ones.

[SJE]

@Wash:
Every person who drives 2 days and rides 3 cancels out a 3 day driver. But I would bet that the majority of cycling trips in DC are done by people who are occasional riders: mode share, riding one or two days, or only riding when the conditions allow.

Thus, you systemically undercount the demand for bike infrastructure. THen you get a ratcheting effect: the belief that only a few percent ride justifies bike unfriendly policies which depresses riding.

IIRC, there were big changes when the census allowed people to list their race as other than white, black, hispanic or asian. Suddenly we find that there is a whole lot more nuance in "race" and there was a lot more interacial mixing. Of course, the mixing happened a long time ago, but the census didnt reflect it.

[washcycle]

But I would bet that the majority of cycling trips in DC are done by people who are occasional riders: mode share, riding one or two days, or only riding when the conditions allow.

Well, I guess that's the question. I don't see a reason to believe that any more than to believe the opposite. Why would you be willing to make that bet?

[Early Man]

The best evidence I have that the Census numbers are pretty good ballpark is that they match up nicely with the 24 hour time use diary collected in the American Time Use Survey. The nice thing about the ATUS is that it is a record of what people actually did the day before they are interviewed. No need to decide on a primary mode of transportation. However, under counts are more likely than over counts because of trip chaining (if you bike to the coffee shop next to work first it might not be counted as a commute). The active commute mode share in the ATUS in 2006-2008 was 4.83% for the 60 largest MSAs and 4.89% for the same cities in the 2009 ACS. I'll provide a cite for this once the manuscript get accepted for publication.

[SJE]

@Wash:
I agree its not known, but I am willing to bet on it.

For example,CaBi's own data shows a lot of last mile-first mile usage paired with Metro: these are people who would not be counted as bike commuters.

According to the census, there are approximately 9000 bike commuters in DC (3% of 306K). This is less than the number of people taking CaBi on peak days.

[washcycle]

Right, that's what I tried to capture above - the combo commuters who use bikes as part of the cocktail. According to the 2012 ACS data there are 13,000+ bike commuters. That's on average. On peak days it is obviously higher. On bad days it is obviously lower.

[turtleshell]

I don't work for NASA and maybe I'm overestimating the quality of satellite photos, but would it be possible to look at some fancy satellite photos and count the cyclists in DC? I know we have a lot of trees, but we must have some experience counting bikes from the sky over in North Korea. Comparing the number of cars not in parking spaces vs. the number of bikes would be an interesting number.

[SJE]

@Wash:
Perhaps we are talking about the same thing.

The census estimates 9000 riders (3%) in 2013, the 2012 ACS data says its more than 13,0000, and early Man's ATUS survey estimates an average of just under 5% "active" in large cities (we can assume DC is higher).

This suggests that the census systematically under-reports and the most likely reason is their methodology.

[washcycle]

NASA images aren't good enough to pick out cyclists, nor are any of the commercially available images that I know of. NGA might have photos of that quality, but they won't be declassified for 50 years or so, and they probably have more images of Arak than they do of DC.

[dynaryder]

I think you meant NSA. NGA just does mapping.

FYI,I can type my Dad's address into Bing maps and the overhead photo clearly shows him in the field on his tractor. So commercial imagery is good enough.

[turtleshell]

I just tried to find bikes on the capital crescent trail and on the Rock Creek trail with Bing and Google. Bing seems a little clearer and the photos were taken when the leaves weren't on the trees, unlike google. I can see a few cyclists, but its really really hard to tell the difference between a cyclist and pedestrian. The photos are kind of useless. How about photos from a plane or drone? Might be tough to get permission in DC, but it might be worth doing in another city. Counting the bikes would be pretty accurate.

[washcycle]

Actually I meant NRO. But NGA does more than map. They analyze satellite images.

http://www.spacenews.com/article/nga-awards-big-satellite-imagery-contracts

[DE]

I may be missing something, but I can't see what information you'd get from satellite photos that you couldn't more accurately get from bike counters.

[washcycle]

DE, I agree. A pretty intensive count was done by some researchers about 10 years ago that determined that about 2% of all vehicles in the L'Enfant city are bikes. That's a much more useful number than the commute rate in my opinion. It takes into account travel time, distance, exposure, not commute trips etc...

Of course such a count is hard to do well. Where you count and when you count will dramatically impact the results.

It would be cool if someone could utilize all the traffic and MPD cameras around to town to automatically determine the bike/car ratio on the roads at all times of day and on every day of the year. That would be much more robust data.