How Much Does Density Matter?

It's been a while since we've had a good chart around here, but this Paul Krugman post got me wondering about the relationship between national population density and national auto dependence. So I took his data on which percentage of trips are undertaken in a car, and put it together with population density information. The relationship winds up looking real but also fairly vague -- speaking English seems to lead to driving a lot in a more clear-cut way than does high population density.

But of course overall density data can be misleading here since some countries contain large empty wilderness areas that reduce density but contain so few people that they don't have a noteworthy causal impact on overall commuting patterns. If you lopped the Yukon Territory, the Northwest Territory, and Nunavit off of Canada, their density would be way higher but the total population would be very similar. One might get a more enlightening comparison by comparing some of these European countries to specific states. New Jersey, for example, is slightly denser than the Netherlands but seems to feature far more driving.

Comments

Your second paragraph is dead on, and no one ever seems to address this.

Posted by LaFollette Progressive | May 12, 2008 12:22 PM

Speaking English is also correlated with lower tax rates, stingier government welfare services, and a tendency toward first-past-the-post voting. Call it a hunch, but I'd bet the last of these has a lot to do with the rest.

Posted by jhupp | May 12, 2008 12:31 PM

That graph might look better if you flipped the axes so that density is the predictor.

Posted by JW | May 12, 2008 12:32 PM

To get an apples to apples comparison, you would have to focus on the urbanized portion of metropolitan areas, down to the level of the census tract or at least the city/township. Counties won't work because some urban counties such as Maricopa, AZ and San Bernardino, CA have lots of non-urban land.

Posted by nodakdude | May 12, 2008 12:35 PM

The interesting thing about Krugman's graph (the one on the link, not Matt's) is about how the Dutch numbers are so low- 28% by bike and only 7% by mass transit (lower than Canada). That is absolutely an issue of density- things are built close together, so biking is an easy thing to do, even when the weather is cold, dark, and raining (i.e. 9 months a year). Outside of the Amsterdam-Utrecht-Rotterdam area (which is, admittedly, where half of the population lives) the public transport in the rural areas of Gelderland or Groningen isn't that different from a similar part of Alabama. Its just that the cities are built in such a way that biking is an easy option.

Posted by Kevin | May 12, 2008 12:39 PM

Some people use a measure of average density weighted by population, which indicates how much density the average inhabitant is experiencing, so the large uninhabited areas don't skew things. Can't remember what the measure was called, though. Didn't you or Ezra blog about it recently?

Posted by KCinDC | May 12, 2008 12:43 PM

As noted above, the second paragraph renders the chart meaningless, and even the apparent correlations are just coincidences. It's certainly something susceptible to study, but it needs a lot more rigor than national population density can provide.

Posted by Rich | May 12, 2008 12:45 PM

You could get far more aggressive with your pruning of Canada: in addition to lopping off the northern territories, you could chop off the top half of BC, Alberta, Sask., Manitoba, Ontario, and Quebec, and hardly lose any of the population. The Canadian population is quite concentrated in urban areas and the southern Ontario peninsula.

Posted by Joe Buck | May 12, 2008 12:51 PM

Lotsa problems here, e.g., 'Percent of trips' is pretty useless without data on the distribution of lengths of trips.

Posted by MattF | May 12, 2008 12:51 PM

I'd concur with that final point, too: It would be far more fruitful to study individual states or regions. This is in effect what Amtrak is already doing in its analysis of the relative success of Northeast Corridor trains compared to longer overland routes. Where the served areas are most dense, the train is a vital alternative to highways, and one with much broader political support than in the nation at large.

Posted by whereisted | May 12, 2008 12:53 PM

That population density for Sweden is likely just as misleading - well maybe not quite - as the Canadian number.

Posted by Aaron | May 12, 2008 1:00 PM

Love that stat on NJ.

If you're going to adjust population density for those areas that are huge but have no population, you also, by similar logic, need to adjust "percent of trips" for total number of trips per person per day. I mean, sure, those Dutch and Swedes use bikes a lot of the time when they leave the house -- but what if that's only because of the greater percentage of Americans who barely leave the house at all? Huh? Dja ever think of that? Huh?

Well, okay, so if that doesn't make any sense at all? There must be some way to make it come out that Americans are better than Swedes on this damn chart. C'mon, let's get a Petroleum Institute of American grant to work on it.

Posted by brooksfoe | May 12, 2008 1:01 PM

As KCinDC implied, basically any density measure that isn't weighted by population is going to end up being more or less useless for these purposes, regardless of the administrative level of the areas being studied (nations, states/provinces, or cities). So, you don't need different units of comparison so much as you need a better measure.

Posted by DTM | May 12, 2008 1:03 PM

I've often wondered if population density might explain a lot of differences between countries, not only transportation but even worldviews. It seems like living in closer proximity to other people would make you more aware of the interdependency of people within society, and therefore more amenable to things like welfare spending or public projects. Not to mention that the tax base for some things would be so broad that each individual's tax burden would be relatively small. Mass transportation is an obvious example--you can only do something like that where your ridership density is high enough to support the operation. Living in a high population density society, you would see that, yes, people can collectively get things done through government, because you would personally see it happen. I'd like to see more data on the subject.

Posted by JimN | May 12, 2008 1:05 PM

If you're looking at metropolitan areas, it would be important to include year of founding. Older cities were built and designed before the automobile, so they're less likely to be designed with cars in mind. That's why the older cities are more bicycle/pedestrian friendly. Back in the day, bicycles, horses and your feet were the only way to get around and horses were expensive.

Posted by Mo | May 12, 2008 1:08 PM

Workplace Density

Workplace density matters too, especially in America where transit options are not so robust. Some medium sized cities in the US with relatively low density have relatively high public transit use percentages. Portland and Pittsburgh pop to mind. Portland has a very nice system to begin with and smart land use policies that provide service from downtown to low and medium density areas. Pittsburgh has geographic challenges galore and no land use policies to speak of but its major economic centers are central and dense thanks to being built pre-cheap oil. Neither of these places have ridership percentages like an overall dense city like DC or NYC, but do better than comparable cities like Charlotte and Columbus that sprawl all over the place willy nilly.

Posted by Roboticghost | May 12, 2008 1:15 PM

New Jersey is comparatively dense, sure, but it's worth remembering that North Jersey is a completely different beast than South Jersey. Most of New Jersey's density comes from the areas surrounding New York, and I think you'd find that there would be a lot greater use of public transportation there because of the large number of people who commute to NYC for work.

That having been said, though, there's still a hell of a lot of driving going on there, and it's largely because (as Matt has hit on before) the US infrastructure is, for the most part, built completely around cars. Even if you live within walking or biking distance of somewhere, it's often logistically almost impossible to actually get there by those methods, making driving more of a necessity than it really should be.

Posted by Fargus | May 12, 2008 1:15 PM

New Jersey is comparatively dense, sure, but it's worth remembering that North Jersey is a completely different beast than South Jersey. Most of New Jersey's density comes from the areas surrounding New York, and I think you'd find that there would be a lot greater use of public transportation there because of the large number of people who commute to NYC for work.

That having been said, though, there's still a hell of a lot of driving going on there, and it's largely because (as Matt has hit on before) the US infrastructure is, for the most part, built completely around cars. Even if you live within walking or biking distance of somewhere, it's often logistically almost impossible to actually get there by those methods, making driving more of a necessity than it really should be.

Posted by Fargus | May 12, 2008 1:17 PM

I think a more salient factor would be the distance from where people live to where they work. My dad, for example, drives about forty miles to work every day in Northeast Ohio. Ralph Regula spent his lengthy Congressional career building highways but no light rail. So, if wanting people to drive less is desirable, then it requires an ambitious national public works and transportation policy or a change in culture where people live closer to where they work, but I really don't know any stats about distance of residence from place of work.

Posted by Matthew Struhar | May 12, 2008 1:17 PM

Another thing to keep in mind is land use patterns. As was noted on the very blog in posts a bit back, many urban areas in the U.S. are designed to discourage foot traffic, and with little alternative to private automobiles when it comes to motorized traffic. Thus even if density were high you'd still see most travel taking place in cars.

Posted by jimBOB | May 12, 2008 1:21 PM

It seems like living in closer proximity to other people would make you more aware of the interdependency of people within society, and therefore more amenable to things like welfare spending or public projects.

Well, that depends. If you look at comparatively dense areas of, say, suburban North Jersey and suburban England, you'll find subtly different patterns of land usage. Some of that's on account of zoning or urban planning requirements: local pubs and shops, footpaths and cyclepaths connecting homes and schools, etc, change the character of neighbourhoods.

When you're using local shops, local pubs, local parks, then you do end up talking to people who live nearby, rather than treating your home as a kind of pod.

It bears repeating, that this isn't about imposing Manhattan on the burbs: if you take a glance at suburban development outside the sprawlier bits of the US, you'll likely see building done very differently. People may not bike to work, but they certainly don't need to tramp along a grassy verge by the roadside to get to the bus stop.

Posted by pseudonymous in nc | May 12, 2008 1:31 PM

If you're looking at metropolitan areas, it would be important to include year of founding

I suspect this will become even more important going forward. A pre-war street grid and rail line presence are huge assets in the era of expensive gas. Its cheaper to move stuff by train than by truck, and water beats 'em all. The Rust Belt's looking shinier. Every thing that made Cleveland, Detroit, Buffalo, and so on attractive places to manufacture things before the ubiquity of cheap oil will benefit those places going forward. Not strictly transit related, just sayin'.

Posted by Roboticghost | May 12, 2008 1:32 PM

"That having been said, though, there's still a hell of a lot of driving going on there, and it's largely because (as Matt has hit on before) the US infrastructure is, for the most part, built completely around cars. Even if you live within walking or biking distance of somewhere, it's often logistically almost impossible to actually get there by those methods, making driving more of a necessity than it really should be.

Posted by Fargus | May 12, 2008 1:17 PM"

Good point. Walking from my childhood home to the nearest train station, only a five minute car ride away, takes a lot longer because you either have to walk on narrow roads with sidewalks while dodging cars going around blind turns or jump over fences. Building more sidewalks would make the US a lot more livable.

Posted by Reality Man | May 12, 2008 1:33 PM

And southern England is far more dense than northern England or Scotland. Try driving anywhere here when school breaks in the summer--you simply can't. You'll end up parked on the motorway for hours. Yet plenty of people still drive, and drive really big cars.

I suspect it has to do somewhat with affluence and disposable income as well.

Posted by KathyF | May 12, 2008 2:31 PM

The three most populous metropolitan areas in Sweden comprise about 37% of all of Sweden's population. The combined populations of the New York, L.A., and Chicago metropolitan areas comprise about 14% of the total U.S. population. The New York metropolitan area would need to more than triple in density to comprise as large a percentage of the U.S. population as Stockholm does for Sweden. Throw in how much of large American cities developed after the car became commonplace, and I suspect that explains much of the prevalence of the automobile in the U.S.

Posted by Will Allen | May 12, 2008 2:45 PM

When discussing density, there is simply not enough emphasis being placed on urban vs. suburban developments. For example, New Jersey is dense, but the vast majority of settlement patterns are suburban. North is the New York suburbs, South is Philly. There are not enough truly urban, large, dense settlements that would be conducive to walking or public transportation. Pennsylavania on the other hand, is less dense, but it has 2 real urban centers that are conducive to modes of transportation other than driving. So in that sense, NJ's density is a subruban annomaly to realy density.

Posted by Roman | May 12, 2008 3:28 PM

The problem is with the way density is measured. The numbers we are going by are simple to calculate, they are population/area. If you translate this into what it means statistically, what it tells you is this: If you were to pick a random square mile of the country, what is the expected number of people living there? If you think about it, this isn't really the question we want answered.

What we want to know is what's known as perceived population density, which answers the following question: If you pick a random person from a country, what is the expected number of other people living within distance X of that person?

Answering this question gives very different results and gets rid of the anomolies of Sweden/Canada being low density even though the vast majority of their popultation lives in densely populated areas. Of course the answer does vary depending on whether your choosing a distance of 1/4 mile or 10 miles or something in between, and this data is harder to calculate, but it is an objective measurement that could be calculated and would provide much better data for this sort of analysis. But I haven't seen this sort of thing calculated anywhere. It should be.

Posted by Eric L | May 12, 2008 3:38 PM

Density may one way to look at it, but aren't there some other things affecting how much people drive that might play larger factors?

For instance, getting a driver's license in Germany used to be quite expensive, so those with lower incomes did not usually drive.

And America is a car culture, with vehicles used partially for practical purposes (getting from A to B), but also for status and pride (e.g. people who go out cruising on Sundays in their hot rods).

In other words, while looking at density and driving patterns is very interesting, it is actually only a small slice of the picture.

Posted by Mark D | May 12, 2008 3:47 PM

Pennsylavania on the other hand, is less dense, but it has 2 real urban centers that are conducive to modes of transportation other than driving.

Big vs Small

PA has 2 major metro areas, but several smaller ones like Wilkes Barre/ Scranton, Erie, Allentown, and Harrisburg that have transit systems and people that use them. Not many, but some do and others easily could. Morgantown, WV has a fantastic transit system that serves the university population. Transit in smaller cities is very doable. Older, smaller cities, at least.

Other than that, what Roman said. NJ has a very different sort of density. It seems to me that much of New Jersey could be better compared to Maryland, Las Vegas, or Phoenix because the development patterns are similar. Suburbs, suburbs, suburbs. Bad situation.

Posted by Roboticghost | May 12, 2008 3:49 PM

First:
Nunavut, not Nunavit

Secondly:
You do not address either climate or geography as factors. Speaking as a person who lives in Nunavut (which is what Nunavummiut means), at -40, bicycling is not feasible.

Speaking as some-one just returned from the Netherlands, it is MUCH easier to bicylce around in the flats of the Dutch countryside than it is through a place like Vancouver, BC.

Posted by Nunavummiut | May 12, 2008 3:56 PM

But Vancouver is *totally* flat. Much flatter than Toronto, in fact, which is often called a flat city (just that the elevation is down, into the valleys and ravines, and not up). It's the mountains around Vancouver where biking would suck.

As for these figures, hmmm. I don't think it's unfair to say that the USA is basically a global outlier on this issue, though I would love to see Australia represented too. Even though Canada, for example, doesn't do hugely better on transit ridership it is still considerably ahead, despite having a heavily suburbanized population outside of the few large urban centers. The difference, though, is that transit and land use planning are very much part of the Canadian political dialogue right now, so much so that even in car-crazy Alberta huge light rail projects are being pushed through. Combine that with the aggressive "Places to Grow" plan in Southern Ontario, which basically says that the place to grow is pretty much existing urban areas, and Canucks start to look comparatively progressive.

Yet those Europeans (and Japanese!) are still, of course, way ahead.

Posted by Matt C | May 12, 2008 4:27 PM

If someone's dense, it usually matters.

And the US's population is pretty dense, considering that thirty percent of them still support Bush and millions more support Clinton.

Posted by Richard Steven Hack | May 12, 2008 9:09 PM

The relationship winds up looking real but also fairly vague -- speaking English seems to lead to driving a lot in a more clear-cut way than does high population density.

That may be so, but the graph is, in fact, designed to obscure any such relationship. If you wanted to get insight into whether population density affected propensity to drive, you ought to set up the graph so that population density was on the x axis, where we normally put the independent variable. Setting up the graph in the way you have it makes it much more difficult to figure out the relationships.

Posted by Sean Peters | May 12, 2008 9:23 PM

I estimated the data points and entered them into the spreadsheet of a statistical package, JMP, and ran a regression. The correlation is a -.504 which is reasonably strong. The nature of the relationship is reflected in the regression equation where the intercept is 65 and the slope is -.025. This means that the predicted percent of trips by car is 65 for the hypothetical country with a 0 density and that for each hundred points that density increases the percent of trips by car declines by 2.5% or for a thousand point increase the percent of trips declines by 25% to 40%. If I knew how to show the graph I would but the plot is not as useful as the statistics.

Posted by jim gundlach | May 12, 2008 11:28 PM