The most recent data collected by the U.S. Census’s American Community Survey (ACS) shows that the number of Portlanders who commute to work by bicycle has dropped from 4.4% in 2006 to 4.2% in 2007.
The survey asked, “How did you usually get to work last week?”, and 10,987 people (out of just over 280,000) replied that they used their bicycle (the 4.2% number is arrived at after subtracting the number of respondents that work at home).
This decline would be the first drop in Portland’s bike-to-work mode split as reported by the ACS since at least 1990 (when the number was 1.12%).
(UPDATE 10/8)Portland still ranks #1 among the 50 U.S. cities with the most workers. The ACS says Portland’s 3.9% bike-to-work rate (they leave in the folks that work at home) is just above Minneapolis at 3.8%. San Francisco (2.5%), Seattle (2.3%), and Tucson (1.9%) round out the top five.
It’s important to note that this survey only accounts for bike-to-work trips, which experts say account for only 1 out of 8 total trips. Jessica Roberts, a former BTA staffer and now a planner with Alta Planning in Portland previously told us that it’s widely accepted that these numbers vastly under-represent mode splits.
According to a story in the Eugene Register-Guard:
Six Oregon cities Eugene, Portland, Bend, Medford, Salem and Hillsboro made the top 100. Portland ranked 14th on the list, and No. 1 among the 50 American cities with the most workers, according to the survey.
Someone I spoke to at PDOT about these results was surprised by the decline and said, “Every other count we have is up.” At yesterday’s Transportation Safety Summit (hosted by PDOT), Commissioner and mayor-elect Sam Adams briefly mentioned that a preliminary count of bike traffic across the four main downtown bridges is set to pass 17,000 trips this year (up from 14,500 in 2007)
To delve further into the ACS numbers and methodology, check out the official website.
[Note from editor: Due to some shoddy reporting, I have gone back and cleaned up this story a bit after getting input from commenters. Thanks.]
Thanks for reading.
BikePortland has served this community with independent community journalism since 2005. We rely on subscriptions from readers like you to survive. Your financial support is vital in keeping this valuable resource alive and well.
Please subscribe today to strengthen and expand our work.
I’d be very surprised if ridership didn’t increase in Portland in 2007, but this doesn’t mean much, it is well within the sampling error. They don’t sample very many people for the ACSs, so they tend to have fairly large margins of errors. But they publish the margin of error:
2006: 11,477 +/- 2,106
2007: 10,987 +/- 1,587
So what they are really saying is that in 2007 there is a 68% chance that the number of people that rode to work was between 9,500 and 12,500, (and a 95% chance that it is between 8,000 and 14,000.)
Just wait for 2008 numbers. They will be huge.
Not a big deal. Cycling is still an extremely popular mode of travel in Portland, and this is only the trip to work portion of it, which generally is only about 25% of people’s weekly travel. There are tons of cycling trips being made that aren’t accounted for in this figure. Many avid cyclists, for one reason or another, don’t ride to work. But they’re riding to the store, to visit friends, to go out for entertainment or a bite to eat. And that’s great. Any trip not made in a car is a plus for the city.
Someone payed attention in their statistics class. =)
Thanks for reminding us all that we should remain skeptical of any statistics, Matthew.
Yeah Matthew! Join the fight against math illiteracy.=)
Well, y’know…Lies, Damned Lies and Statistics…
polls are only as good as people taking them…I commute every day all year long and I didn’t take the poll…
Polls can be kind of like weather forecasters who predict a possibility of showers, when if they’d just look out the window they could see it’s raining cats and dogs.
I see lots of people on bikes. Seems like more than a couple years ago.
One other question I have around this survey is does it account for population growth? If the population grew by more than 0.2% between the 2 sureveys then the gross total of bikers on the road actually went up.
Matthew, good catch, but the ACS “margin of error” represents a 90% confidence interval (+/- 5% margin of error), not a 68% one (cite at the end of this post). You’re still correct, though, the 2006 and 2007 estimates are not significantly different at the 10 percent level. The decline wouldn’t generally be considered meaningful.
Jonathan, the statement “it’s widely accepted that these numbers vastly under-represent mode splits” doesn’t make any sense. Mode split represents some fraction of the total (100%), so by definition if some mode splits are underestimated others must be overestimated. It can’t be the case that all modes are “vastly underrepresented.” I imagine what you mean to say is that certain mode splits (e.g. bicycling) tend to be underestimated in the ACS? Maybe Jessica can explain further.
From the ACS: “Data are based on a sample and are subject to sampling variability. The degree of uncertainty for an estimate arising from sampling variability is represented through the use of a margin of error. The value shown here is the 90 percent margin of error. The margin of error can be interpreted roughly as providing a 90 percent probability that the interval defined by the estimate minus the margin of error and the estimate plus the margin of error (the lower and upper confidence bounds) contains the true value. In addition to sampling variability, the ACS estimates are subject to nonsampling error (for a discussion of nonsampling variability, see Accuracy of the Data). The effect of nonsampling error is not represented in these tables.”
Yeah, I’d go so far as to say that since the results are within the margin of error, the headline is false. The survey says that the fraction of bikers remained the same.
What week did they use?
Was it during that run of great weather we had during the summer, or was it recently when the weather went to crap and the seasonal riders mostly drove?
83% of statistics are completely made up. 🙂
Does this mean that the fixie fad is over? Have the hipsters moved to vintage Vespas or the most ironic choice of public transit?
Brad..oh god, please, can we only hope…
Kt @#12:
I heard it was 80%, but you are still within the margin of error of 100%. ;o)
I’m just wondering how many *new people* arrived in Portland in 2007 from California, Vermont, Russia, Vietnam, etc. without bikes and/or a clue how to get around Portland yet.
The survey is done is January or February, so the survey only captures the people that commute by bicycle year round. The bridge counts are done in the summer, so they’ll tend to capture the fair weather only crowd as well…
How about a shout out for Eugene. Number two in the study with an 8.5%. Impressive for a city the League of American Cyclists rates as only a Silver. I guess there is more than one way to do it.
The problem with many surveys these days is the proliferation of cell phones. Since it is illegal to contact individuals on their cell phones for a survey, the only people who respond are home phone owning people. They did do mail and in person. They also try by mail, but my guess is they get a low return in the by mail part.
Dear God, the thread has been hijacked by math geeks – the is the single most awesome day of my life!
I think more than anything this illustrates that you can’t accept any numbers at face value until you know the source of the data and the collection methodology – unfortunately not every agency publishes that information when they publish their numbers. And the average joe focuses on the number, and doesn’t care about the methodology (if it’s published, it *must* be true!)
Matthew Denton wrote:
“The survey is done is January or February, so the survey only captures the people that commute by bicycle year round.”
I’m not a Census expert, but I’m pretty certain that the ACS survey is continuous (check http://www.census.gov/acs/www/Downloads/tp67.pdf, p. 7-2). It looks like they mail out to a different portion of the sample each month, and then respondents have 3 months to return the survey.
The old knock on the decennial Census with regard to bike mode share was that it was mailed out in the spring “shoulder” season when weather is really variable a lot of places. The ACS estimate should reflect the average bike commute share across the seasons. There’s still the problem of having to choose 1 mode to represent 5 (or more) days worth of mode choices).
Hey didn’t by chance do there survey the during the dead of winter when ice was slathered everywhere?
I’m sure the is methodology to correct for this, but asking what I did in the last week is not often representative of what I did in the week before.
i ride everywhere all the time, but i don’t even have a job! go figure.
I must have been riding my bike when they called to survey me…
“90% of all statistics can be made to say anything 50% of the time.”
Direct TV commercial
Here in Minneapolis, with their ACS results showing them a close second to pdx in bike commuters, I’ve received numerous emails touting that “we’re bridging the gap to Portland!!” which I find amusing, encouraging, and threatening all at the same time. 🙂
Here in Minneapolis, with their ACS results showing them a close second to pdx in bike commuters, I’ve received numerous emails touting that “we’re bridging the gap to Portland!!” which I find amusing, encouraging, and threatening all at the same time. 🙂
Okay, you are right, (again.) The Jan/Feb date that I saw in the accuracy document refers to when the study is done at certain types of group quarters, (prisons, nursing homes, etc…)
Well, when the margin of error exceeds the 0.2 % change, I’d say those numbers don’t give a lot of inormation. At the same time, why aren’t the numbers growing more rapidly? I mean, I would have expected a huge jump so no change or a possible slight drop, feels like a big drop. In a way these numbers represent the feelings being expressed on here a couple months ago when Portland didn’t even get into the top 3 bike friendly cities. People were frustrated that change wasn’t happening faster. It seems like everyone is patting themselves on the back rationalizing the meaninglessness of these statistics when maybe we should be asking why it’s not increasing more rapidly? Have we maxed out with our current roadways? Is there resistance? Are the die-hard commuters making newcomers feel unwelcome? Are drivers revolting and making newcomers feel unsafe? Is it unsafe with our current traffic laws? Does bicycling feel to many like a counterculture statement?
It is true that statistics CAN be deceiving and you have to know how to read them, but at the same time, they aren’t actually meaningless as many like to proclaim and we wouldn’t use them if they didn’t give us valuable information. Whether there was a drop or not, there wasn’t a significant increase, and that’s what’s troubling.
One more thing, from my perspective seeing more bicycles is not as important as seeing fewer cars. I’d really like to see those numbers if they are collected. I do feel like there is less traffic than there used to be. Is it just my imagination?
>>>>Coyote: How about a shout out for Eugene. Number two in the study with an 8.5%. Impressive for a city the League of American Cyclists rates as only a Silver. I guess there is more than one way to do it.>>>>
The easy way for a town to get a big mode share for bicycling is to have a college or university with an attendance level that’s a significant portion of the local population.
I’m not a Census expert, but I sometimes play one at work. To clear things up:
#19- You’re right- the ACS is a rolling/continuous survey throughout the year (250,000/month nationally)
#9- The number of surveys mailed out does account for population growth (roughly 1 per 500 people- yeah, a painfully small sample).
Like others have said- there’s no statistical difference between 4.4% and 4.2%. Given the range of error surrounding the numbers, that’s putting waaayyyy too fine a point on it.
-Amy
Since we’re talking about bike trips to work, maybe there should be some accounting for the rise in unemployment…?
hey folks.. .thanks for all the input on this article. I should have just presented the data and let you all do the reporting on it.
apologies for not spending more time on this… i rushed it up in the wee hours this morning. I’ll go back and do some clean-up of the story if I can.
Wilsonville= remote island! LOL
Ian- unemployment declined between 2006 and 2007.
Eileen, #29, while I agree with your point, the census is a very course measurement tool, so all I’m saying we shouldn’t put too much weight in what it says. It is like trying to measure how think a piece of paper is, with a yardstick. If the piece of paper is twice as thick as it was last year, it is very hard to tell that with a yardstick. And we have better tools to measure it, for instance, Bridge/Intersection counts which are based on a count of actual bicycles on the ground, (so have much smaller errors,) and the city auditor report, which asks about primary and secondary transportation modes, which captures people that only ride to work a day or two a week…
And in any case, we should be trying to increase mode share faster, even if it is increasing already. 🙂
I’m afraid the new headline is also misleading. Just goes to show that interpreting statistical data is hard. Thanks for reporting on this survey, Jonathan. It’s good information to have. However…
The problem with the old headline was that the data are not good enough to indicate a shift, since the change is well within the margin of error.
The problem with the new headline is that it uses the number from 2006 instead of the more current, lower number from 2007.
I think the best way to handle statistics like these is to include the margin of error when reporting the number. So we’d have 4.4% +/- 0.8% for 2006 and 4.2% +/- 0.6% for 2007. (those are 90% confidence intervals, by the way, according to the ACS site)
We might get a better idea of the true values for these numbers if we combine this survey data with some other data, for example the numbers for the bike counters on the Hawthorne bridge at typical commute times. If we saw that those numbers did increase, we might suspect that the 2006 survey overreported the bike mode share, while the 2007 survey underreported.
I certainly don’t expect a detailed statistical analysis here, but I would appreciate indications of the margin of error and confidence levels when reporting on surveys like this.
Phew. Math dork mode off. Time for a beer. 😉
There are a number of known problems with the ACS. It’s bad at forecasting total numbers (as in, the total population of a city, or the total number of bicyclists in a city). For this reason, if the total population of Portland went up, and it is extrapolating from its sample to attempt to tell you the total number of bicyclists in Portland who commuted to work… it’ll get it wrong.
All the ACS really says is that, of the sample of people that they asked in Portland, fewer rode bikes in 2007 than in 2006. Keep in mind the following:
* The ACS only samples 250,000 people per month — across the ENTIRE UNITED STATES!! (population, what, about 300 million?) In theory, this leads to 3 million samples a year, or 1/100th of the population. Over a 10-year period, that equals 1/10th of the population, which is what the old Census Long Form used to sample (the ACS is the replacement for the Long Form, which will not be collected in 2010).
So… since the ACS is sampled continuously, every month, you can imagine that if they happened to sample different neighborhoods in say, November through March of 2007 than in 2006, they could EASILY come up with a .3% difference in the results of how many folks were bicycle commuting.
As Matthew basically points out, the difference is thus NOT statistically significant.
“The problem with the new headline is that it uses the number from 2006 instead of the more current, lower number from 2007.”
argh.. my mistake Dag. I’ve corrected the headline. I think next time a survey like this comes out, I’ll just email a few of you “math geeks” and ask if you’d do the story for me ;-). thanks again.
Put another way…
At the 1/100 samples per year and Portland’s population of 537,081 we’re talking 5,371 samples for Portland. A change of 0.2% is only 11 samples. So the 0.2% change means that 11 people answered that question differently from one year to the next, out of the 5371 who responded for all of Portland. On the other hand, the error range of +/-0.6% means that 33 more (or fewer) people could have answered the question a different way. 0.2% is not statistically significant.
I was one of the responders on this survey. Even when I was taking it I had to wonder where the bias was. The questions were skewed towards drivers over bicyclists and totally missed any other form of transportation.
Once again I get to resort to 3rd grade remarks…
Statistics Schmatistics!
While stats may be over used by many, they are taken with a grain of salt by me.
I am on the streets, and what I see is this.
“More commuters than ever before.”
I don’t need to government to sway my opinion on that with their stats.
Coyote (18) and PdxMark (31)-
Thanks for mentioning Eugene- we did come in number two (after Boulder) and are pretty proud of the achievement and though the University certainly has a lot of riders we also have a large community of folks off campus who ride.
I also think this survey has some skewed numbers and measurements but it does show us a change which we all see on the street level.
Also, regarding our BF city status…. we are silver but haven’t reapplied for a status change in a couple years and hopefully with the next application we’ll get up to Gold at least.
Anyway, for once we come ahead of Portland in something bike-wise… well for once in the last few years- it used to be all about us- really.
Mark, the one thing I remember from statistics class (I was more of a number theory gal) is that if the sample size is truly random, 5,000 is actually a huge sample size. Don’t quote me on that because I am remembering back 15 years to a class I didn’t study that hard for. I DO remember that my professor rode his bike to work!=)
Oh! Just found this excellent sample size calculator for you all:
http://www.raosoft.com/samplesize.html
fun times
The ACS data can also tell us the about race, ethnicity, and income-level of people who commute by bike or by foot and you can do this fairly easily using the website that Jonathan cites above.
I remember playing around with 2006 ACS data for Multnomah County. If I recall correctly the data indicated that bike and pedestrian commuters tended to be white but also lower to middle income. That kind of fits the younger, white demographic, I tend to see on the streets but it would be interesting to see if this is changing.
Jim Labbe
Watch out, Minneapolis is working hard.