Thursday, September 29, 2011

Grease in the numbers

Numbers are messy.  I know we like to treat numbers and precise and concrete, but it just isn't the case.  So you know I just have to react to this factoid in the news.  Via the PG and BikePgh I see the headline factoid that bicycle commuting is up 269% in Pittsburgh.

To be clear, I wish it were true that bicycle commuting was up that much.  I think Pittsburgh has a great bike history, a lot of great bike amenities, has had some great bike leadership, and I have no doubt a great bike future.   But +269%?  Maybe.. but also possibly maybe not.

The literal number of bicycle commuters for the City of Pittsburgh reported in the 2010 American Commutny Survey (ACS) was 2,267. The margin of error reported with the ACS on that specific number was +/- 824. Big range there. Again, the large relative error is due to the small size of the city and the really small size of bicycle commuters.  The ACS is a sample frame of roughly one in 100 households in the general population annually.. higher oversampling happens for smaller groups or smaller geographic areas. So the 2010 data the reported 2,267 bicycle commuters living in the City of Pittsburgh likely comes from the survey capturing just a couple dozen bicycle commuters. Not a lot to work with.

So what has the ACS been telling us year in and year out?  Here are all of the comparable ACS results on bicycle commuting of City of Pittsburgh residents. ACS results before 2006 did not include Group Quarters population and so I would not consider them comparable to the others.  Shown are just the ranges made up of the estimates plus or minus the reported standard errors. The ACS reports different data based on compilations of data over 1, 3 or 5 years.  The 3 and 5 year compilations have more data and thus  smaller proportional sample errors than the 1 year estimates.

So a lot of confusion over any one point estimate is the fault of the Census Bureau resulting from the elimination of the long form of the Decennial Census.  The result is that we are all going to be forced to rely on data from the American Community Survey(ACS) for most all data like this.  The problem is that as a survey, there are limitations that are not always appreciated.  The message the Census Bureau tells a lot of folks like me to spread is that you have to take into account that smaller areas and in particular smaller populations can have results significantly impacted by sample error and other issues.  Bicycle riders are a small population for sure.

Clearly there was a big jump for sure from 2000 when the reported number of bicycle commuters was 627.  In absolute numbers that works out to a percentage increase far larger than +269%.  Not clear to me from what I have read on what the 269% refers to specifically actually.  The actual increase could actually be higher depending on what is being measured.

This is all a good learning exercise on how to use ACS data which will be coming out nearly continuously from now on.  If you think that the number of bike riders is relatively stable, or at least changing in a consistent way then reconsider the graph above. Reality probably does not jump around all that much.  So yes, all of those ranges show an increase over 2000.   When that growth happened is another question altogether.  If you look at the high end of the range in 2007 and the low end in 2010, it is at least conceivable that the number of bicycle commuters has gone down over that time. 

Over the decade though, the number of post-secondary students in the city proper has gone up a lot.  Those folks are disproportionately non-auto drivers.   If students report that they are working as many are even if part time, they will be asked their mode of commuting and in most cases their response will be included in the results for commuting data.  With something on the order of 10K more undergraduate or graduate students living in the city since 2000, I have to believe they are a big chunk of the increase in bicycle commuters showing up in the data.


Anonymous MH said...

It seems like I see more bikes than I did a few years ago, but maybe that's because the bike lanes make them easier to see.

Thursday, September 29, 2011 10:11:00 AM  
Anonymous Anonymous said...

Hey, I was biking to work in Pittsburgh when there were only 3 of us, give or take 2.

In other news, last week SPC had a young man stationed at the intersection of Darlington Road and Schenley Drive, at the top of the golf course, counting bikers heading into Oakland.

Thursday, September 29, 2011 10:36:00 AM  
Blogger erok said...

i agree with your points, although i do think that these acs numbers are good as a comparable to other cities and to show trends.

I disagree with your point in the post that the grad students increase is the reason for the cycling increase. if that was the case, you'd also see an increase in the walking mode share, which we don't see. I've been cycling in town for over 15 years, and anecdotally, more people of all ages/demographics are riding to events, the grocery store, work, coffee shops, etc. Some of the same events that me and a handful of my friends would be the only ones riding. the same bike racks that were once empty downtown, are now full or over capacity. i could go on.

Thursday, September 29, 2011 12:19:00 PM  
Blogger C. Briem said...

When I get a chance I'll generate the cross tab of post-secondary students and employment/journey to work to show the point. May not work for the city proper because the sample is so small.. but more broadly it will likely show up.

a different point.. but I've been watching gender stats in transit and biking for some time. Hard to use most of it. Look at the data direct from the League which includes that. Look at the variability in share women as bike commuters. There are places showing up at 100% women bicycle commuters.

Thursday, September 29, 2011 12:25:00 PM  
Blogger C. Briem said...

and just curious if you know.. but what numbers are generating the 269% calculation? As I said, I actually get a higher number.

Thursday, September 29, 2011 12:26:00 PM  
Anonymous BrianTH said...

Isn't the standard answer to stick with one of the rolling averages (maybe three-years for now, five-years when it is available for more years) and track that trend? Still some error you'd have to account for (so fit a line and report the fit), but it would smooth out a good chunk of the artificial year-to-year volatility.

And I'm guessing if you looked just at 6-8, 7-9, and 8-10, you'd see a pretty steady trend (only 8-10 is missing from the chart, and it seems clear rolling off 7 and adding 10 should result in another increase).

Thursday, September 29, 2011 12:27:00 PM  
Anonymous MH said...

On erok's point, I don't see why you'd need to see an increase in the share of people walking. There are only so many houses you can talk to Oakland from and, aside from some new dorms, that number is fairly constant over time. What with the cost of parking and the crowding on the buses, I'd guess that most people who live within a passable walk have been walking for years.

Thursday, September 29, 2011 1:26:00 PM  
Blogger erok said...

" but what numbers are generating the 269% calculation?"
That was direct from the League's post and calculations.

"I don't see why you'd need to see an increase in the share of people walking"

I'm basing this off of this comment " Those folks are disproportionately non-auto drivers."

Thursday, September 29, 2011 3:46:00 PM  
Blogger C. Briem said...

ah.. it is the change in proportion between 2000 and 2010 expressed as a percentage change. Not quite the best way to present it, but I get it. Like calculating the percentage change of a percentage. Can be confusing. 0.44% to 1.63% works out to the +269% kind of sort of.

Thursday, September 29, 2011 9:02:00 PM  
Blogger Chris said...

This data also oversimplifies everything. I am a bike commuter - I ride my folding bike to the train, take the bike on the train, and then ride to work, but this survey counts me as a transit commuter. I think it's ridiculous, but there you have unaccounted for cyclist. Maybe there are more like me?

Monday, October 17, 2011 4:02:00 PM  
Blogger C. Briem said...

I'm not sure we have any such train commuting here in Pittsburgh. :-(

Monday, October 17, 2011 4:04:00 PM  

Post a Comment

<< Home