"In fact, the U.S. Census Bureau estimates that approximately half of all Americans live within 5 miles of their workplace."
The web page is advocacy (yay) for biking to work. Unfortunately, the bicycling advocacy community has some echo-chamber characteristics. It's entirely possible that Mikael, the CDC and a couple other people are essentially all quoting each other.
The journey-to-work information collected on the US Census is in the form of an address for the workplace, time of travel (both when and how long) and mode of travel (and if more than one mode, you're supposed to say the mode for the most distance, so if you bike to a bus, you took the bus).
While it is _possible_ the US Census took the address for the individual filling out the form, the address on the journey-to-work section, and did some calculation on it (either crow-flies or some google-maps style get directions), that strikes me as far, far, far beyond unlikely. And if they _did_ do that, I cannot imagine they wouldn't have produced something I could stumble across somewhere.
But hey, I'm wrong a lot. Prove it to me. I've e-mailed the CDC asking them for sources on the above quoted statement. Bet anything the net effect of asking is to cause the bike to work page to be edited and/or retired. Or the US Census quits getting "credit" for this statistic.
ETA: Okay, judging by this:
What we're looking for is someone doing O/D (origin destination) block analysis. The above was done for Minnesota, with a median commute distance of 6.23 miles. Which is _not_ less than 5 miles. Last time I checked. Furthermore, it is _not_ 8 km or less than 8 km. It is 10 km. And furthermore, the difference between 8 km and 10 km is greater than the daily average cycling distance in Denmark, the country with the highest annual cycling average in the world. So don't be telling me 2 km doesn't matter.
ETAYA: Further digging supports the crow-flies analysis theory.
I don't think the CTPP O/D pairs pass through the addresses for obvious reasons, so detailed route analysis, in addition to being computationally expensive, may be impossible.
Again, supporting my assertion that median commute distance in miles is greater than Mikael's assertion in km.
ETABlah blah bleeping blah: I don't know if you've read down this far, but have you noticed how everyone is throwing out commutes longer than 100 miles one way?
ETABBBB2: On the, maybe Mikael means North America theory, here's something:
median commuting distance in Toronto in 2006 was 9.4 km. NOT under 8 km.
I'm feeling a little cranky, actually.
But for Canada as a whole it 2006, median was 7.6. Which _is_ under 8 km.
Wow. Look at that chart! It's the crazy rural areas that bring the numbers down. The cities are considerably higher.