Did NYC Bike Commuting Decrease in 2009? That’s What The Census Says

According to the American Community Survey, the share of NYC commuters using their bikes as their primary mode decreased from 2007 to 2009. Many are skeptical, however. Graphic: Noah Kazis
According to the American Community Survey, the share of commuters biking as their primary mode decreased from 2007 to 2009. Many are skeptical, however. Source: League of American Bicyclists/U.S. Census ##http://www.bikeleague.org/resources/reports/pdfs/2009_bike_2000_2009.pdf##PDF## Graphic: Noah Kazis

The Census Bureau released the numbers from the 2009 American Community Survey earlier this week, offering a detailed look at how Americans get to work. As Angie noted on the Streetsblog Network, some unexpected cities like New Orleans and Honolulu jumped up the chart on bike commute rates. Here in New York City, believe it or not, the ACS showed a small decline in the percentage of commuters who bike to work and a drop in the total number of bike commuters.

This marks the second year in a row that the ACS has recorded a decrease in NYC bike commute rates, which doesn’t jibe with other measurements of city bike volumes and contradicts the widespread perception that bicycling is increasing. The NYC Department of Transportation screenline count recorded a 26 percent year-over-year increase in cycling in 2009, and Transportation Alternatives put the increase at 28 percent. So what’s going on?

The divergence between Census data and those two measures of bicycling is not new, and points to shortcomings in each method. On the one hand, the Census almost certainly underestimates the number of cyclists, by only counting people who bike as their primary mode of getting to work. That leaves out people who only bike a few times a week, or only for errands or recreation. On the other hand, DOT’s screenline count measures cycling into the Manhattan CBD, which can distort the picture if cycling rates are much higher in Lower Manhattan and northwest Brooklyn than the rest of the city.

New York won’t have a clear and reliable estimate of citywide cycling until more reliable metrics are released. For now, the numbers are the subject of some dispute.

Rutgers professor John Pucher, author of the study “Cycling in New York City: Innovation at the Urban Frontier” [PDF], argued that the ACS data are more reliable as a citywide metric than DOT’s count. “I trust professionally trained statisticians who try to create a statistically representative sample as opposed to screenline counts which cannot claim to be representative in any way,” he said.

But the ACS findings include a few hard-to-swallow stats that have some questioning the results altogether. “I just don’t see that any credence can be given to these numbers,” said Charles Komanoff, who developed TA’s estimates. He pointed not only to the widespread observation that more bikes than ever are on the street, but to inconsistencies in the borough-level ACS data. For example, the ACS shows the number of bike commuters in Staten Island falling by almost 90 percent. “586 out of Staten Island’s previously counted 658 bike commuters would have had to die, retire, get thrown out of work, move or give up cycling,” scoffed Komanoff.

Similarly, TA’s Noah Budnick argued that DOT’s counts seem more accurate. “Just look at the streets,” he said. “They’re teeming with more bike riders than ever before.”

Michelle Ernst, an analyst with the Tri-State Transportation Campaign, also suggested that the borough-level ACS data were suspect, saying that the sample sizes were simply too small to be accurate. She wasn’t willing to dismiss the ACS data entirely, however. Instead, she noted its limitations, wondering whether overall cycling trips may be increasing while bike commuting stagnates.

Even if the borough-level numbers are extremely imprecise, however, the divergence they present is striking. While the ACS showed plummeting bike commuting rates in Staten Island and the Bronx, it recorded an increase in Brooklyn. That lends some credence to those who say that DOT’s screenline counts capture a western Brooklyn-based upward cycling trend more than a citywide one.

“That screenline count is totally unrepresentative of what’s going on in the city as a whole,” said Pucher. “It’s not impossible that in that area of west Brooklyn and the Lower East Side, that you could still have that 26 percent increase in the screenline counts” even as bike commuting decreased citywide. Pucher said it’s telling that the ACS doesn’t diverge from Portland’s local counts, which he said are more thorough than New York’s and include a travel survey.

Pucher said he didn’t know why citywide bike commuting might have declined, but that with two consecutive years of ACS data saying so, he’s convinced.

To get a better sense of what’s happening on the street, many are anxiously awaiting two forthcoming data sets. Ernst highlighted the 2010 Census, which will have a much larger sample size than other studies. “We’ll get much more detailed data, block-level data,” said Michelle Ernst. “Anecdotally, I’ve seen a lot more people out on their bikes, in the bicycle lanes, and especially a lot more women,” she continued. “Maybe this is happening on a very local level.”

Pucher said that he’s excited to see the results of this year’s National Household Travel Survey. The release of that information has been delayed to make some technical revisions, he said, but once released it will help paint a much fuller picture of American travel habits. “It’s a fairly large sample and it includes both work trips and non-work trips,” he explained.

Until the new data is released, we have two imperfect and contradictory sets of information — and our own observations. Let us know what you’ve seen on the streets of New York. Is cycling up across the city or just in certain parts? Bronx and Staten Island readers, especially, does the Census data match what you’re seeing on the street?

  • Perhaps the falling numbers can be explained by rising unemployment. It’s hard to commute if you don’t have a job to go to.

  • Noah Kazis

    Jonathan: It’s certainly the case that the recession is responsible for at least some of the absolute decline in the ACS numbers. But that doesn’t explain the decline in mode share: bike commuting declined from 2008-2009 more than driving, walking, or taking transit. It also doesn’t help you square the ACS with the DOT estimates.

  • Clarence Eckerson Jr.

    I gotta say without having statistical numbers to prove it – I have been riding my bike for almost 20 years in and around NYC. This year it has been incredible the number of cyclists riding. Recently, I have even started using the word “crowded” to describe some bike facilities during rush hour. We are not Copenhagen, but it feels a little like Portland and San Francisco these days riding the bridges, separated lanes, and the Hudson bike path.

    Of course, statisticians probably don’t want to go near anecdotal evidence, but I know I am not the only one. We need a true survey – more of what Portland has been doing for years.

    On Wednesday, I stopped at a favorite pizza joint at Flushing & Washington by the Navy Yard. It was 4 PM, not even rush hour. In about a minute, I saw a dozen people using the Flushing bike facilities in both directions, two cyclists riding out of Steiner Studios, and a handful turning off/turning on Washington. I just smiled.

  • All census operations are flawed in that they are not really a census, but merely a survey, usually a manual one, executed for a limited time.

    This made sense when this was the only way to acquire data. However, automation scales better.

    In Copenhagen and now Helsinki, traffic counters at the end of bike paths count hourly, daily and total traffic, both for the display of passers buy and for later analysis.

    in a world with near-universal cellphone possession amongst adults, scanning for discoverable bluetooth devices has been show to pick up one adult in six, so generate datasets for pedestrian and bicyclist movements in a city -over many, many months.

    Here, for an example, is the statistics on pedestrian traffic near my house over a six month period
    http://bristolcars.blogspot.com/2008/12/datamining-streets.html

    where you have multiple scanning points -such as in the city of Bath-, it becomes easy to distinguish traffic modes by their speed between sample stations, and so build up an accurate map of traffic movements round a city.

    As I said: surveys and censuses have to viewed as a historical best practise, a substitute for modern instrumentation and data mining

  • J:Lai

    I have a been a regular biker in NYC for many years, although at no time would have identified myself as a bike commuter. I make about 10-15 subway trips per week, and about an equal number of bicycle trips.

    Most of my bike trips are around northwestern queens and northern brooklyn (areas that have very poor transit connections.) Over the last few years, I have seen a dramatic increase in the number of bike riders. This is especially true on the roads forming a network of bike lanes.

    Side question: is there a good way to travel south from Flushing Ave through clinton hill/prosp hts? I usually take Washington Ave, but wish there was a better route.

  • Doug

    Of course bikers lose their jobs at a higher rate than other commuters: we sweat more and are also more self-righteous. Who would you get rid of?

    Of course, car drivers are self-important and transit riders also get stuck in the heat of un-air conditioned platforms. But those are issues New Yorkers have been dealing with forever.

  • J

    I think I’m going to reserve judgment for the 2010 census, which actually asks a decent sample of residents how they get around. I have a hard time seriously believing that 4 times as many people in the Bronx commute PRIMARILY by sreetcar than by bike (3,431 vs 736). The only streetcar in the region is in Jersey city, and it takes a minimum of 45 minutes and two other modes just to get there from the Bronx. What does Mr. Pucher have to say about that? Does he believe it, just because a trained statistician says so?

    Statistics can be very useful, but they need to be tempered by logic and observations.

  • J

    Also J:Lai,
    You can try the new Vanderbilt Ave bike lane/sharrows. It’s not bad and tells motorists exactly what they need to do to pass bikes. It also has good connections at both ends. Enjoy!

  • J:Lai

    Thanks, J!

  • Bashing the Census when it doesn’t show the answers you think it “should” is a very old game.

    Something to keep in mind is that the ACS is a statistical survey with a margin of error. You can’t make statistical conclusions from the data just by looking at the headline number.

  • JamesR

    I see fellow bike commuters of both the folding and full-size variety on my Metro North reverse commute all the time. Who is counting us? I highly doubt MTA has stats on us and NYCDOT’s screenline count doesn’t hit us. It’s probably the same thing for those using LIRR and NJT.

  • Larry Littlefield

    “I think I’m going to reserve judgment for the 2010 census, which actually asks a decent sample of residents how they get around.”

    The 2010 Census will ask no such questions.

    As a compromise with the Census Bashers, the Bureau agreed to eliminate the Census Long Form, previously sent to one in six households, which provided detailed social and economic data. The census just asks a handful of questions now.

    As part of the deal, the Bureau replaced the long for with the American Community Survey, which is an annual survey with a much smaller sample size — and therefore less reliable data for phenomena with small numbers such as bike commuting.

    There isn’t enough of a sample size to report census tract data for any one year. The Bureau plans to release tract data for the period from 2005 to 2010 on average.

    Predictably, some Republicans now want to get rid of the ACS, and with it all data on income, poverty, educational attainment, occupation, industry, ancestry, household type, etc. etc. etc. And commuting. What you don’t know won’t hurt you, I guess.

  • Ian Turner

    Aaron is right. The graph at the top of this article should have error bars.

  • Jeff Arp

    What Larry said.

  • @Aaron — what makes you think I didn’t look at the error bars in the ACS data? Of course I did. The 89% drop in S.I. bike commuting falls many standard deviations beyond the means. Ditto for 58% in Bronx.

  • J

    Thanks Larry,

    Did those changes take place for the 2010 census? Was there no long form sent out?

    Also, perhaps I was just bashing the census for not showing what I thought I’d see, but I really think biking is growing here.

    Finally, what I really want to see is the NYC screenline counts for 2010. Anyone know when those will be released?

  • Larry Littlefield

    “Did those changes take place for the 2010 census? Was there no long form sent out?”

    Yes, no long form in 2010. The American Community Survey started with a pilot in 2000 along with the last long form, and has been expanding. The first census tract data is supposed to be for 2010, but it won’t be for 2010, it will be for 2005 to 2010 to get enough sample size.

    Perhaps the ACS is missing young people and immigrants, thus less biking. Or perhaps young people and immigrants are figuring out that New York is of, by and for the seniors, and leaving — though that’s not what I see on the street.

    It’s nice getting annual data, but you pay a price. I don’t know who came up with the bicycle numbers. The ACS data that is published lumps bikes in with other things, where with the long form there was enough sample size for more detail. Perhaps someone used Public Use Microdata Sample Data and didn’t worry too much about sample size.

    If they don’t have at least 50 cyclists in the survey for each place they report, the data isn’t valid. NYC would have to pay the census bureau to do a more detailed survey, as it does for the Housing and Vacancy Survey.

  • @Charles,

    My comment was a reaction to Noah’s graphic in at the top of the story, which shows a very small decline YoY, only 0.03%. It isn’t hard to imagine how that could be statistical noise.

    I did pull the numbers on Staten Island, however. They are just tiny:

    2000: 364
    2006: 97
    2007: 157
    2008: 658
    2009: 72

    Those are total number of bike commuters, btw. To put this in perspective, 67,289 people (+/- 5,152) used public transit to get to work.

    Looking at this, 2008 is clearly the anomaly, not 2009. The others all fall within the margin of error. Even if the survey somehow missed 600 people, in a city of 8 million, that’s a blip.

  • Amy

    I work on various college campuses and have been commuting to the north Bronx (University Heights/Riverdale/Kingsbridge) for several years now. 4 or 5 years ago I was the only person using the lone campus bike rack. This year, that same rack typically has 3 or 4 bikes locked to it each day, and the college has already installed additional racks, with plans for even more. And at another Bronx college I have noticed an even greater increase and even seen days when the (albeit small) bike racks are full. I have also noticed a definite increase in the number of people using the Harlem River bike path (which desperately needs repairs and better maintenance) and the number of cyclists crossing the University Heights bridge at Fordham Road. Although in total, the number of people cycling in the Bronx may still be relatively small, it would appear that there has been a significant relative increase in the last few years.

  • I agree with Aaron, these numbers are still small enough that if they *didn’t* fluctuate wildly it would be odd.

    When we get things to about 15-20% bike commuters (what I think is a reasonable goal.) a change of 600 will hardly show up at all.

  • I wonder what is the proportion of responders in 2007 who were in the business of making deliveries?

  • J

    I think the overall gist of what I’m reading (and thinking) is that both the sample size and the actual number of bikers are both small enough that wild fluctuations are to be expected, which is why the census gives a relatively wide margin of error. On the whole, while higher numbers would have been encouraging, this should not be seen as failure or even a setback. It should remind us that there is still a LOT of work to be done to move biking from a fringe activity to a normal part of life.

    In terms of bike commuting, we’re not going to see a dramatic increase until the cycle tracks really link the bridges with the hearts of the office districts. The Manhattan Bridge-Allen St.-1st/2nd Ave connection is a good start, and it will be interesting to see how it changes things as people discover it. However, most office jobs are north of 34th Street and the 2nd Ave cycle track still has gaps. Worse, the Queensboro Bridge still dumps you into bike hell, and the 8th & 9th Ave cycle tracks still have poor or no connections in Midtown. It’s not surprising, then, that commute numbers are still low. Each segment of cycle track certainly helps, but it is only when big sections are linked that people start changing their habits. When it does happen, though, the momentum will build quickly and the census will clearly show it.

  • Larry Littlefield

    “I think the overall gist of what I’m reading (and thinking) is that both the sample size and the actual number of bikers are both small enough that wild fluctuations are to be expected, which is why the census gives a relatively wide margin of error.”

    Again, the city is paying the Census Bureau to do a more detailed survey every three years, the Housing and Vacancy survey. Its purpose is to prove the city’s vacancy rate is less than 5.0%, thus justifying the perpetuation of a “housing emergency” and the continuation of rent regulations.

    If the city wanted a more accurate version of anything, is could pay more and add questions to this survey.

    For example, a more accurate measure of crime could be to ask everyone about victimizations in the past three years.

    Transportation could be asked about — not just means of transportation to work, but otherwise.

    HVS income data is notoriously bad, because so much of the city has income-tests for preferred housing (vacancy decontrol) so when people hear what it is for they slant accordingly.

    But there is already “neighborhood conditions” data in the survey.

    The only problem with serving other purposes is, HPD wants to control the survey to “make sure” the vacancy rate is under 5.0%.

    http://www.nyc.gov/html/hpd/html/pr/vacancy.shtml

  • Larry Littlefield

    I would go so far as to say that transportation is the only normal Census data item the HVS doesn’t ask about. They’ve got employment, occupation, industry, income by source, education, race, ethnicity — in addition to all the questions about housing.

    And the next one is in 2011.

    A few other question could provide more accurate data on transportation, and perhaps other things. Like do all the city’s educated workers come from elsewhere, because the city’s schools have been so bad.

    http://www.census.gov/hhes/www/housing/nychvs/2008/questionnaire.pdf

  • I think Jonathan’s comment at the very top deserves a little more credit than it’s getting. The decrease in all travel due to unemployment certainly can explain the relative drop in cycling mode share, because now all the other modes of traveling are less crowded and thus more desirable.

    The same can be said for the reduction in gas prices due to the economy. There was a noticeable uptick in transit use, carpooling, and cycling when gas prices were much higher and unemployment was much lower, so it shouldn’t be much of a surprise that cycling would take a bit of a hit in the current circumstances.