Hacker News Reader

Comment by davidw | Hacker News Reader

I really don't get what this is supposed to represent, exactly.

No one commutes from Ashland, Oregon to Portland Oregon - it's a 4+ hour drive. Yet they're in the same pink area. Meanwhile, no one really commutes from Bend to Portland either (2.5 hour drive in good weather), but we're brown over here.

There's certainly some more cultural affinity between cities in Oregon's Willamette Valley than with the more conservative central and eastern part of the state, but even that only holds for the larger cities - there are plenty of pretty 'red' places down the I-5 corridor.

View on HN

The algorithm was a clustering algorithm, and takes as input the number of clusters desired. The researchers put in 50 clusters, so at some point the algorithm declines to separate sparse regions, because the separations are better used elsewhere.

I wonder if there's a way to... Not specify a number of clusters? Maybe use some sort of commute threshold.

K means can also solve for k, but its a more opened ended and difficult problem. This is pretty good approximation

It's designed to identify economic regions. The way they do that is to graph all the connections between cities and cut the least-connected ones until 50 clusters emerge.

It doesn't, as you say, mean that someone on the edge of a graph commutes all the way to center---after all the drive time for some of these routes is more than 8 hours.

You could argue that choosing 50 clusters is arbitrary and it would make more sense to severe connections below some threshold "quality" and see how many clusters meet that criteria.

Yeah, I don't understand the point or value of these groupings either. San Antonio and Austin are really completely separate economic entities with little economic overlap. There are a few Austin commuters who live a bit more south than some San Antonio commuters, by that doesn't really "join" the areas economically.

I can't find a closer image of the west Portland area, so I can't comment on the nature of those commutes.

I live in SF, and at work my team consists of ~16 people. Of those 16 people only 3 live in SF (city/county). The other 13 live in the East Bay or south of the city. Of those who live outside the city the majority doesn't get paid enough to live within city limits; the others get paid well but have families and couldn't reasonably afford a large enough home in the city.

People are very willing to convert time into savings (on housing). And sometimes they have no other choice.

> People are very willing to convert time into savings (on housing). And sometimes they have no other choice.

May be, but they aren't commuting from Chico! That's a 3 hour drive in good traffic.

I have worked with people who commuted from Sac to Oakland or SF a few days a week.

edit: Original post said Sac, not Chico...

I'd like to see a closeup of the Portland area too. I live in Portland and while there's been a tremendous growth in the community over the last 10-20 years, Oregon has the concept of "urban growth boundaries" to preserve farm land, especially small family farming.

So most of the population here is within around 20 mi of the center of Portland, unlike Phoenix AZ where it sprawls out 100 mi or more. I'm guessing there are people who travel from/to nearby regions (e.g., Newberg, McMinnville, the "wine country") for commerce, tourism, and probably some commuters.

Perhaps there are similar considerations in other regions like SF which certainly has "tentacles" that stretch a long way out there.

I think part of this is due to selecting "50 areas" to divide the country into. They mention that in the article and even say "it has a math basis" but don't explain why. If they chose 60 (for example) some of these areas might naturally divide. It might be a very interesting thing to look at deciding how many boundaries come about by applying limits to either economy or population to try to balance each.

An even crazier example is the Arizona Strip -- that is, the part of Arizona north of the Grand Canyon -- being clustered with the rest of Arizona.

You have to drive 200 miles out of the way to get from one side of the Grand Canyon to the other. Since the northern Strip is very sparsely populated and relatively poor, the number of people there who have the means to regularly commute by plane or helicopter must be zero.

The region north of the Grand Canyon should really be clustered with Utah or Nevada, not with Phoenix and Tucson.

In the article (graphic 5/5) it seems to be part of the Las Vegas megaregion.

It's not that everyone from one region commutes to the center every day, or that everyone inside this region shares a culture.

Use the number of commuters as a metric for 'closeness' of two areas. Then take it further by asking "how 'close' is Portland and Ashland?" by using the sum of the 'closeness' of each little hop from area to area.

This map is showing clustering, finding groups that are nearest to each other, using that metric as a distance function. Find me 50 areas where '# of commutes' are the distance function, and you get this map.

Ashland and Portland aren't very close geographically, but in terms of economics, they're in the same neighborhood of the country.

Oh i'm sure people commute. A 4 hour drive is fairly reasonable. I live in Denver and commute to Kansas City (8.5 hours) once a month to visit my folks. If you include gas prices and the price of rentals it comes out to be cheaper than flying and renting a car in KC. With shorter distances this would probably be a huge factor for many. For example, most folks in KC drive to StL which is a 3.5 hour drive. I don't think anyone I've ever known in KC flew to StL.

When the article says "commute" it's mainly referring to people who commute daily (or often) for work.

There are people who commute for clusters of days, though. If you work a three-on, four-off schedule, commuting in to start the three and out when the three are done is perfectly reasonable for a distant job and doesn't preclude having a healthy home life, too. There are also people who are mostly remote and only come in once a week for face-to-face meetings.

Airline pilots can commute from Phoenix to Minneapolis regularly.

When I think of a commute, I think of a daily activity that takes someone back and forth to work. Given that definition it seems totally unreasonable for someone to spend 4 hours going one way.

A small number of people do stuff like this.

I worked with a consultant who lived near Scranton so he could commute to a large number of places. (Syracuse, Albany, NYC, Jersey, Philly)

Drove 3 hours a day one way!

The dataset includes things like NYC-LA, so it clearly includes non-daily commutes as well.

The final clustering excluded trips between points >=262 km from each other though.

> A 4 hour drive is fairly reasonable

Erm. No, it's not. I'm sure there are a tiny percentage of people doing this out of necessity, but spending 8 hours a day commuting is not even remotely 'reasonable'.

Reply on news.ycombinator.com