Tuesday, January 17, 2012

The population density to support my ass

I swear if I read one more time that such-and-such a place "doesn't have the population density to support" transit, or sidewalks, or high speed rail, or multifamily housing, or mixed-use development, or sushi restaurants, or anything other than Joel Kotkin's suburban American fantasy, I'm gonna spit. At this point all you urbanists, livable streets advocates and transit supporters really ought to know better. Yes, you, Richard Layman, but you're not alone so don't feel bad; you're just the most recent person to say it, who happened to set me off.

Layman is summarizing the myopic "lessons learned" in an article about Spain's high-speed rail system (don't miss the "next page" buttons!) by mainstream reporter Tim Sheehan of the Fresno Bee, based in part on an interview with political economist Germà Bel. Sheehan writes:
"There is no question whether (Spain's system) can cover its costs. It cannot," Bel said. "It actually has not recovered one single euro from the infrastructure investment. The government claims they are recovering the operating costs, but the numbers are not clear."

The busiest high-speed lines in the world are capable of making money, Bel said, including between Paris and Lyon, where about 25 million people ride the French TGV trains each year, and the Japanese Shinkansen trains between Tokyo and Osaka, which draw about 130 million riders a year.

"But this is not the case with any single line in Spain," Bel said. "The most crowded operation is Madrid-Barcelona, and it has not even had 6 million people in a year."
Robert Cruickshank took issue with Bel's argument and Sheehan's portrayal of it. He also faults Sheehan for holding highways to a double standard. What about Spain's highways? Are they half empty too? Do they pay for themselves, since many of them are tolled?

Note that Bel doesn't say anything about population density; I think that's something that Layman is reading into it. What he and the other analysts tell Sheehan is that the system doesn't have high enough ridership, which is not the same thing. In fact, Bel's own research fails to show that population density is a significant factor in urban transit ridership.

In 2010, Bel and his colleague Daniel Albalate did an amazing factor analysis (PDF of the 45 cities in the Mobility in Cities Database (yours for only $1,608 plus shipping!) and found that the factors that loaded most heavily on service demand were as follows:
PRIVATE_TIME (average time spent by private vehicle trip)0.747
FLEET (fleet of vehicles available for public transport purposes)0.553
MOTOR (number of private vehicles per capita)-0.524
PRICE (average price charged to urban transport users)-0.466
PARKING (number of parking spaces per thousand jobs in the CBD)-0.221
DENS (urban population density)-0.190
PUBSPEED (average speed of public transport vehicles in operation)0.0722
Here's a quick note on how to read a factor analysis. A positive value means a positive correlation, so the higher the GDP, the higher the demand for transit. A negative value means a negative correlation, so the more private vehicles per thousand residents, the lower the demand for transit.

The key here is the absolute value of the loadings. The greater that value, the stronger the relationship between the factor and the dependent variable. In this case, the effect of GDP is more than ten times as much as the effect of transit speed. Note in particular that population density and transit vehicle speed have the loadings with the lowest absolute value.

Finally, remember that this shows correlation, not causation. It may well be that high transit demand causes large numbers of transit vehicles to be available (funny how that works!) and high population density, not the other way around.

Again, this article by Albalate and Bel refers to urban transit, not intercity rail, but the dynamics involved in intercity mode choice are similar to urban mode choice. The time difference (3.5 hours for AVE vs. 6 hours by private car), number of vehicles per capita, tolls and gas would all make a difference. Those are some lessons that California can learn from Spain, but as Cruickshank argues, Sheehan was predisposed to find problems with high-speed rail. Maybe Bel tried to tell him, but he wasn't listening for it?


Anonymous said...

Your opening sentence is conflating a lot of things.

By definition, no density is required to make multifamily housing or mixed-use development viable. Every rural family farm is a mixed-use plot (assuming the farmers live there). And even in the smallest town, nothing stops you from building a multi-unit building except for pointless zoning restrictions and money.

High-speed rail doesn't require density any more than airports do. People want to get places quickly, and they'll happily drive (or take transit) to a distant train stop if it will speed up their journey.

But local/rapid transit, in the form of local buses and streetcars and metros, is almost purpose-built to solve the problem of too many people trying to travel in too small a space. They are fast only to the extent that they avoid traffic -- it's rare for any of these vehicles to approach highway speeds, let alone exceed them.

So given the choice of building local/rapid transit in a city, where space is at a premium, or in the suburbs, where there's more arterial and highway capacity than you can shake a stick at, I think there's a clear correct answer.

Yokota Fritz said...

This past week has been full of editorials and articles attacking CA HSR. I wonder if Southwest Airlines is putting a bug into the ears of newspaper editors again like they did 20 years ago in Texas.

Richard Layman said...


Well, Steve Belmont leans more to your position, than mine.


He argues that areas with lower density than I think can support more frequent transit.

But observation finds that in the most dense places, transit works a lot better and carries more people.

FWIW, I can't think of any time that I've argued that urban design and sidewalks are dependent on density (their success is, but that's a different question).

It is true that with bike infrastructure (and transit actually), I argue that it's better to start by putting it in the places it's most likely to be used (based on population density, presence of activity centers, etc.) as it's better to lead from success than failure.

Alon said...

In these difficult times, we should recall Gary Barnes' paper on perceived density. Barnes only studies US cities, and finds a very small positive correlation between standard urban density and transit use, and a somewhat larger positive correlation between perceived density and transit use. But the strongest correlation, and the one that persists the best after pruning New York from the sample, is between the ratio of perceived to standard density and transit use.

That said, the density ratio correlation is probably causal in the other direction: transit use causes spikiness, rather than the reverse. Cities that start building out their transit systems (New York and Paris in 1900, Tokyo in 1910, Hong Kong in 1970, etc.) tend to be dense all over; the spikiness comes only later, as transit transforms development.

alai said...

One thing I've become curious about, which I haven't seen good data on, are the following (and it might be interesting to see if there's a correlation to transit):

The amount of area, as a fraction of land area, devoted to parking, to streets, and to structures (allowing it to go above 100% in the case of multistory buildings). I've read that in the case of parking, for example, downtown LA clocks in at ~80% (making ample use of garages, obviously), while SF is ~30% and NYC is ~15% (this is probably going to line up with spaces/1000 jobs pretty closely).

Another is, in keeping with post in praise of narrow streets, the fraction of land area overall which is devoted to streets. I recall reading somewhere that one remarkable difference between Paris and DC (and other American cities) is that the amount of land devoted to streets is more than twice as large-- something like 40% vs 20%.

I think it would be interesting to see these figures presented together-- building area, street area, park area, perhaps private open-space area, and car parking area, all expressed as a fraction of land area. It seems to me that it could be a pretty good way of comparing different types of urbanized areas in a way that simple population density fails at.