From National Household Transportation Survey data, here are two charts showing how household demographics affect driving habits.
First, there’s income:
Clearly, as income rises, people drive more…but only up to a point! As incomes rise past roughly the US average, driving flattens out. Which makes sense when you think about it: if your salary magically doubled, would you really want to spend more of your time behind the wheel?
This chart may offer a partial explanation for the declines in per capita vehicle travel we’ve been seeing in recent years. Over the past decade or so, incomes for middle class and poor folks have either been stagnant or declining; only the well-off have seen a significant benefit from “economic growth.” But if wealthy folks don’t drive more when they earn more, then today’s skewed version of “economic growth” might not do much to stimulate additional driving.
And here’s a second chart, looking at vehicle travel by residential density.
What this chart shows is clear: all else being equal, the denser your neighborhood, the less you drive. Having jobs, stores, and services close by can shorten car trips, and lets some people walk, bike, or take the bus instead of driving. And as neighborhood density rises, some households will find that they don’t need a car at all. (Note: I restricted this chart to households with annual incomes of at least $50,000, to neutralize any of the income-related effects identified in the first chart.)
Academics have studied the relationship between driving and density exhaustively—perhaps too exhaustively, since many spend an awful lot of time quibbling about the specifics. But in broad strokes, the NHTS data confirms what should by now be an undeniable fact: folks who live in compact neighborhoods simply drive less.
John Niles
Clark,
I don’t think you’ve demonstrated that there is no income effect in that second graph just by pulling out the low earners. In other words, within the range of $50k and up, there may be important interactions between income, residential choices, and driving that are obscured.
Also, how much of America (how many households) does each of those bars in the second graphic represent?
It’s not news that people who choose to live in denser areas drive less. What’s less understood is who chooses to live that way and how many of them are there? The people I know who live in Belltown are young, without children, often without partners, and make a lot of money.
Can the National Transportation Survey you dug into be teased to reveal the effect on driving of all the factors it measures in some sort of multi-variable way? Clusters of income, family size, and housing size?
Clark Williams-Derry
John —
I totally agreed about your first point — the graph is suggestive, but I’d need to do more research to develop a solid model that reflects the relationship(s) among income, driving, and other demographic factors.
But even then, I might not be able to factor in the “self-selection” effect — namely, the fact that people who don’t like to drive often choose to live in neighborhoods where they don’t have to. Figuring out the size of the self-selection effect is important, but almost certainly requires a different set of data — looking at how attitudes affect driving, and/or at how moving to a neighborhood with a different density affects driving habits. And you’d probably have to look at more than just density, since there are plenty of other neighborhood factors that affect driving; density is usually just a proxy for them.
But unfortunately, the NHTS microdata doesn’t seem to be available yet. So I’m stuck using their data extraction tool — which means, for now, no clever modeling based on individual survey responses. (To be honest, that’s probably out of my comfort zone anyway…)
As for the carless–obviously, very few fit the “Belltown hipster” demographic. Nationally, most are poor. But it is possible to use the data extraction tool to look at specific slices of data, holding household size and income constant. For example, if you look at all households with just 1 member, you see roughly the same pattern: a wealth effect (where driving goes up to roughly middle-income, then flattens–though with more noise in the numbers!) and a density effect (where driving goes down as density increases). You see the exact same thing in 3-person households, though with different numbers: miles per driver increases in 3-person households.
But to do it right, you’d need the microdata. Which I don’t have…
Jon Scholes
John (and Clark),
Here’s demographic data on Downtown Seattle that answers some of your questions and shows that density isn’t just for the rich and single. Two important findings here, the median Downtown household income is lower than the citywide median, racial diversity is greater than what is found citywide, and over the last two decades we’ve seen significant growth in the number of children residing Downtown.
http://downtownseattle.org/files/file/Demographics2011_WEB.pdf
John Niles
Jon, interesting data, thank you.
Unfortunately, I don’t think I’m seeing cross-tabs and time trends in this data that provide clues to help us decide whether people who have the resources to own a car are going to keep on living in denser places as they grow older, get better jobs, and have families. The policy goal is to suppress driving, right? Walk, bike, take the bus instead.
We know that denser residential living suppresses driving. The question is, how are we doing as a society with our green-growth policies in motivating the car-owning segment of society (most people) to keep on residing and walking in the denser communities — like downtown Seattle and Capitol Hill — where they don’t need to drive so much?
Since the comparison of the 2000 Census to 2010 Census showed a shift to the suburbs across America, I’m guessing we’re not doing very well.
Did Western Washington show overall densification over 2000-2010? I know Seattle became more populous, and King County grew within the UGA but what about Pierce, Snohomish, and Skagit?
There’s a map of dots showing where all the Microsoft employees of the region live in the pdf http://www.leg.wa.gov/JTC/Documents/Studies/TransitAdvisoryPanel/Microsoft_JimStanton.pdf. Lots of dense residential areas are inhabited by this workforce, and some not so dense. And how will it look in five years?
And with driving, trends over time are important when controlling for density. Much of the drop over time right now is from a bad economy since 2008 and historically high gas prices. But what about other influences? I just reviewed an academic paper that using travel diary data found cell phone owners exhibited larger geographic activity spaces in which they moved around over the course of a week.
Cass Martinez
In well-to-do neighborhoods, traffic comes IN to service the homeowner: the house cleaners, the gardener, the pool man, UPS, Fed-Ex…
lise goss
True, but the only group in the upper chart who would hire in-service are in the last bar (greater than 100k). Clark, it would be really nice if on that same upper chart there was a second set of data (like a line) that shows the percent of the total population within that group. Granted it doens’t add a lot to the overall results, but it would be good context.
Cave Johnson
This misses what I suspect is the main driver behind the densit/vmt relationship. It’s not mode choice, its trip distance. The more density, the closer everything is together. The closer things are, the less you travel to get to them. Even if you drive, your trip is shorter. This is consistent with the literature that says that centrality is more powerful than density in predicting VMT. A high density subdivision on the urban fringe performs more like a low-denity subdivions on the urban fringe than it does like a neighborhood of comparable density closer to the urban core.
Clark Williams-Derry
True ’nuff. In practice, density, centrality, land use mix, and parking availability/cost all tend to be correlated — neighborhoods near city or town centers tend to be denser; to have shorter distances to stores, services, and transit; and to have conditions where parking is inconvenient or costly. But as I understand things, it’s the correlates of density, rather than density per se, that actually does the “work” of reducing driving.