Wednesday, October 30, 2013

All things transit

Pittsburgh City Paper covers the very very notional subway map once designed by a graduate student some years ago: Former grad student's proposed transit map goes viral  

Of course the harder news is some analysis showing that the city of Pittsburgh ranks high on the percentage of commuters who do not use a car here. See Atlantic Cities: The U.S. Cities Where the Fewest Commuters Get to Work By Car. You will see the city of Pittsburgh ranks high on the list of cities where commuters travel by foot or bicycle.

I really thought I saw the thesis in the past, but given just how pie in the sky it is, I don't recall even mentioning it here. Given that just one small piece of that would require actually building the Spine Line, which remains a nonstarter, the whole concept is a bit ethereal. It really is a story on the power of presentation and that one map. Otherwise this would all remain unnoticed like the scads of theses written on say Maglev over the years.  Let alone the longer bibliography of transit studies here.

The bike story is a bit curious of late.  Note the PG's 30 year retrospective recently touted how little bicycle riding there was here 30 years ago."Thirty years ago, about the only people on bicycles in Pittsburgh were messengers and a smattering of gung-ho pedalers seemingly oblivious to our unfriendly terrain, weather and traffic."  Big change right?  Of course the Post-Gazette's own tumbr news archive recently pointed out the opposite story of how much bicycle riding was growing here though the 70s.  Funny reading both articles together.

For the bike/ped commuting story.  Be clear, I am a big fan of all the progress in making Pittsburgh a better bike place. Still, it is worth taking note that the statistics are about the city of Pittsburgh compared to other cities.  If you did a similar benchmarking against regions, or even big counties, the high ranking for Pittsburgh evaporates.  The city of Pittsburgh is a relatively small part of the region, compared to the the proportional size of other cities in their respective regions.  So anything that is impacted by density, i.e. walking to work, is going to be a skewed number for Pittsburgh. Works for statistics as diverse as poverty, but most certainly impacts the proportion of folks walking or biking to work.

If you want to look at detailed maps of where people bike and walk to work in Pittsburgh, we had this in the last 3 pages of this newsletter. You may note just how concentrated the pattern of non-car commuting even within the city of Pittsburgh.  Once you get outside of the city limits, it makes little sense even to map it out.

Then there is this measurement issue I've gone through in the past. I'll update this graphic comparing all the American Community Survey (ACS) estimates ever produced for bicycle commuting in the city of Pittsburgh. It is a good demonstration of the pitfalls of the ACS because the proportion of bicycle commuters is so low that it highlights some of the problems of overinterpreting ACS survey-based data in general.  This graphic has the estimates for the number of bicycle commuters displayed plus or minus the margin of error that you should use in interpreting the point estimate.  Makes for a different perspective at least.


Anonymous MH said...

Why did the error bar get so big in 2012? Did the get cheap on the number of respondents or something?

Wednesday, October 30, 2013 9:47:00 AM  
Blogger C. Briem said...

In general, the 1-year estimates are larger than the 3-year estimates which are larger than the 5-year estimates. Longer time = more data.

Wednesday, October 30, 2013 9:48:00 AM  
Anonymous MH said...

Sure, but the 2012 error bar is by far the largest error bar of all the one-year estimates.

Wednesday, October 30, 2013 9:53:00 AM  
Blogger C. Briem said...

Back of the envelope... ACS collects data on 1-2%/year. So a point estimate of 2000 is really expanded from 20-40 respondents who said they commute by bicycle. If a year captures just a few more, or a few less bikers in a year you can see variation in their error calculation.. which is not strictly sample error either.

Wednesday, October 30, 2013 10:03:00 AM  
Anonymous MH said...

Oh, right. The error bar for the percentage of cyclists would be about the same, but the error bar for the number would be wider, given the same sampling method. Thanks.

Wednesday, October 30, 2013 1:04:00 PM  

Post a Comment

Links to this post:

Create a Link

<< Home