Total Vehicle Miles Traveled for the past ten years, annualised.

The St Louis Federal Reserve has a pretty cool data graphing tool on their website, and I recently learned from Matt Yglesias that they now have vehicle miles traveled (VMT) statistics for the last 40-odd years. I knew per-capita VMT has been down for a few years, even pre-dating the recession. Some of the reasons are explored in this article, so I won’t drill into them in too much detail here. But I was surprised to find that total VMT is also down, even though population has increased. I’ve put a couple more graphs below the fold if you are curious.

The change in total vehicle miles traveled, annualised.
Per capita vehicle miles travelled, annualised. Note this graph has a thirty-year window.

48 Replies to “Vehicle Miles Traveled Statistics”

  1. So it looks like the middle of last decade was where Americans turned the corner on VMT’s, which drive many of the projections our transportation planners rely on to power their crystal balls. This is especially true for gas tax revenues, and public transit ridership that forecast X amount of increasing congestion, resulting in Y amount of lost time for drivers to drive the models. The corner was turned prior to the great recession of 2008.
    Anyone want to speculate on WHY (gas price, education, living green, etc) or WHAT is the likely outcome of a society that travels less (assuming the trend continues, as the changes to date are really quite small).

    1. When I was growing up the suburbs where empty most of the day. Dads went to work and took the cars. Moms stayed at home. Kids played in the yard. I used to take bike trips on Long Island and between 9 and 5 the boulevards were like 3-land cycle tracks because no one was on them!

      Then came the real growth of the suburbs…instead of just houses there were more malls, more corporate office parks.

      I mean, look at Redmond. In 1990, when I was here, I used to get up in Seattle, jump in my car, and have 520 all to myself as I commuted to Redmond. Well, that can’t happen any more…but back then Redmond, the town consisted of one street with a bar on it.

      Now it is a complete exurb. There is less reason for someone who wants good food to go farther than Redmond Towne Centre.

      1. By definition, Redmond is not an exurb. People commute *to* it for employment. It’s another productive city in our polycentric urban region. On the flipside, Maple Valley and Sammamish would be closer to the exurb definition. But exurbs tend to be large lot, low-density areas. Think Lake Sawyer and its environs. That is exurban. Awful, awful place at that.

      2. It’s funny because before I used “exurb” I was trying to come up with a name for it. Suburbicity, subcity… I have also used the term Agriurb or subag to denote places like Fort Collins which are small cities that grew up within agricultural regions, not metro areas.

      3. Yeah, “exurb” means the complete opposite of what you were trying to use it to mean.

        I believe the term you were looking for was edge city (though that better describes Renton or Tukwila) or “satellite city”.

      4. Exurb basically means the outer ring of suburbs built up after 1990. So everything outside of Edmonds – Lynnwood – Bothell – Redmond – Bellevue – Renton – Kent. Not sure about Federal Way and Issaquah. Outside that, Auburn, Tacoma, Puyallup, and Everett were basically separate job markets: not many people commuted from there to Seattle. Boeing Everett skewed this, however, because people were transferred willy-nilly between Renton and Everett and they had to commute or quit their job.

        Boeing caused a lot of the sprawl, first by being located in out-of-the way places, and second by making it impossible for people to live where they worked because they would suddenly be transferred and have to commute thirty miles. So people gave up living near work and just lived wherever they wanted to.

      5. There is so much confusion over these terms that it’s hardly worth arguing over.

        I think this is a nice set of terms:

        City = The entire geographically contiguous economic unit, regardless of municipal boundaries, development style and history, etc. Notably, Bailo and others are right to point out that the geographical continuity of an economic unit is less important than it used to be.

        Suburb = A sub-part of a city. This is a common definition outside the US, and taking “sub” to mean “sub-part” rather than “subservient” prevents pejorative uses like “Ballard is a suburb of Redmond” or “Boston is a suburb of New York” or “The US will soon be a suburb of China”.

        Exurb = Developing or new suburb; development at the frontier of urban sprawl.

        But there’s more than one way to do it.

      6. That’s a little tight. I would never consider Everett an exurb, though I would consider Marysville one, as with most of the other towns on Highway 9. Issaquah is iffy; in many ways it still feels like a small town that’s not actually part of the metro area, but it’s tied in a lot tighter with the rest of the area than some other towns of comparable size. Sammamish, though, is definitely an exurb, and the Highlands would be too if not for the deliberateness of its planning. I wouldn’t consider Auburn or Federal Way to be exurbs, though I’ve barely been in either, but Tacoma is the only part of Pierce County that’s definitely not an exurb.

      7. Tacoma is the only part of Pierce County that’s definitely not an exurb.

        Lakewood, University Place, Fircrest, Steilacoom?

      8. But that’s not how it works at all. Exurb isn’t “developing”. It is by definition sprawl. It means “extra-urban”, essentially that awkward mistake of land less than suburban low-density SF6 (6 units per acre). It’s in the realm of SF2 to RA-5. It’s completely uneconomical residential large-lot development. I suppose if you want to be unkind to places like Maple Valley and Black Diamond, you can call them exurbs because they entirely dependent upon far away urban centres for their $$ while giving nothing in return, even though they are more dense. But, you can’t redefine the reality–Lakewood and Fircrest do not qualify as exurbs.

      9. Part of the confusion is that there is so much connotative baggage. You cannot have a reasonable discussion once you start calling something “sprawl” or saying it is “unsustainable” without every defining what that means or when it will finally end up consuming itself in a fireball because it can not longer sustain itself.

        I wish rather there were some nice maps of charts of the country which had multidimensional variables like population density, square footage per person, building height…maybe a heat map would start to show just how continuous the whole thing is rather than these artificial divisions.

        Speaking of which, one thing I do question are the old political boundaries which were made back when there were completely discrete “cities” and yet persist into this day. Are these really the governing spheres that we want to participate in? Are some people getting too little representation…too much? And so on…

      10. Puyallup, Lakewood, and Fircrest are suburbs of Tacoma. Puyallup’s exurban-ness lies in how much it has become a bedroom community for Seattle, and how much the city has expanded with acres of new tract housing with Seattle commuting at least partly in mind. It really comes down to whether Puyallipites are more oriented to Puyallup and Tacoma, or to Seattle. That I don’t know, but an outsider’s perception is that the number of Seattle commuters living in Puyallup has increased dramatically.

        Lakewood I wouldn’t call an exurb because it’s more oriented toward Tacoma and Fort Lewis. And maybe I’m biased because my family is from there and the people I know are oriented toward Pierce County, not Seattle. But Spanaway has all the characteristics of being an exurb of Tacoma.

    2. I think there are a number of reasons. One is gas prices. I think the era of cheap gas is over. We had some really cheap gas in the nineties (which played a part in the economic boom we had) but I don’t think anyone believes those days will ever return. Most of the gas we get now is gathered from really expensive places (tar sands, wells that need to be fracked, etc.). If demand slips, then those places just stop pumping oil, and the price bounces back up. Add the fact that it isn’t the U. S. that is primarily driving demand (it is Asia) and no one believes we will ever see $1.00 a gallon gas again.

      Cities are also a lot more attractive. Seattle has always been fairly attractive, but it is a weird city. It’s inner city has never been that bad, so its been harder to view the trend. Oakland is probably more typical. It used to be a slum, now it is just part of wealthy and desirable East Bay. New York city (proper) is also a lot more desirable. A big part of this is that crime rates have dropped like crazy. This has happened all over, but its effect has been felt way more in the city, than the suburbs. People always assumed that suburbs were pretty safe, now they can assume the same thing about cities.

      Of course, there is also the increased number of transit systems. Once you build a rail system, it is built. Transit systems (especially those on their own line) scale really well. Freeways don’t. There are still plenty of people who move to the suburbs, but they don’t have any illusions to being able to quickly drive anywhere during rush hour. They don’t care, as long as they can get to their bus or train quickly.

      Although sprawl continues, it has probably slowed because the suburbs are already built out. Again, Seattle is unusual in that it is still growing very fast. But even in Seattle you can find plenty of cheap housing in the suburbs. I’m sure some folks want places that are farther and farther out, but there are still plenty of existing places that are available in the suburbs, if that is what you want (or can afford). Many of the existing places have adequate transit, thus reducing the overall amount of travel.

    3. The article nailed a lot of it: people are no longer buying the idea that a car is the sign of American freedom and prosperity, the epitome of convenience, and the only way to pick up chicks. There’s also volatile gas prices and the Internet. People who earlier would have been wandering around bored are playing video games and chatting on Facebook and reading Wikipedia.

      What’s the likely outcome? If the trend keeps getting bigger, eventually it will have a decisive effect at the ballot box like gay marriage. Then politicians will turn around or be replaced. It’s not just the young: the elderly are also expanding and becoming too feeble to drive. There’s a counterargument that young’uns may rediscover the suburbs when they settle down and have kids, but even if that happens, the fact that they’re starting from a less car-centric base suggests they won’t be as tolerant of ped/bike/transit unfriendly policies as their parents were.

      I’d like to say the built environment will just evolve with the attitudes and there will be plenty of places like Denmark. But I’m afraid there’ll be a gap where people want alternatives but the alternatives won’t be ready yet and they’ll be too expensive to accelerate.

      What happens to the “red states”, it’s harder to say. Right now they appear to be doubling down on ideology. “Car = freedom, bus = socialism, train = boondoggle.” But with increasing numbers of elderly and young adults, you’d have to guess they’ll become more evenly divided on non-automobile infrastructure and walkability. And they may innovate with local- and state-led initiatives like California has, since that’s more palpable to them than federal.

    1. It could be suburban infill of retail services plus the flight of employers out of the city to where people live.

      1. Hopefully the suburbs will infill walkability too, and become more like streetcar suburbs. Then they can be part of the solution rather than part of the problem.

      2. People live in the city, too. In general, way more people live in a particular city than in a particular suburb. Moving to Bellevue makes sense if you want to serve the folks on the East Side, but if you want to serve the folks in the greater Seattle area (which includes folks in the north and south end) then locating your business in Seattle makes a lot more sense. I don’t know if there have been studies to look at the impact of the suburbanization of businesses (which seemed to increase greatly in the 1980’s). I would guess that it lead to an increase in driving and commute times, not a decrease. For every happy employee that drove a short distance (or even walked) from their suburban home to the office park (AKA campus) there were there was one who had to drive from a different suburb and another one who had to drive from the city. In both cases the public transportation system was inadequate. Even now, it is much easier to get on a bus and travel downtown from the suburbs in the morning than it is to do the opposite. When these businesses started, the reverse commute was very fast. But as the businesses increased, the traffic got worse, and the location less desirable. As a result, I think there has been a trend (by many businesses) to move back to the city, in response to their workers. Amazon is a great example of this.

        Suburban retail infill probably has played a part. For every worker who has to get out to the ‘burbs to work at the Lynnwood Safeway (and I’m sure many live in Lynnwood) there are a dozen folks in Lynnwood who are happy they don’t have to drive to Edmonds just to go grocery shopping. Retail services are duplicated everywhere (unlike employee centers) so locating them in more places reduces travel. In other words, people will go their local Safeway, but they will still travel across town to get to Microsoft headquarters.

        Suburban infill in general may have played a part. It is quite possible (I haven’t looked at the numbers) that the country as a whole is not sprawling as fast as it once was. This would make sense. Lynnwood has a very different feel than Seattle. If that is what you like, then there are lots of places like that. The greater Seattle area has huge amounts of suburban land that is available. You can still buy a new house on a new lot, but it will probably look a lot like an “old” one. There are lots and lots of those “old” ones. That wasn’t true thirty years ago.

  2. The difference between this financial crisis and normal recessions is that employment has not returned to healthy levels. It would be great to find a trend for lower VMT, but depressing if that trend is predicated on a new economy with permanent unemployment on this scale. Optimists are looking for a culture change among young people for explanation, but I’d keep unemployment in mind too!

    1. I’m sure unemployment plays some part in this, but the drop happened before the recession. For King County, this trend started as far back as 2000. And remember, Andrew’s charts don’t take into account population growth, meaning per-capita car use has dropped more than the charts show. Sightline has had an excellent series looking into this phenomenon.

  3. I looked at the graphs, and thought “Wow, that’s a huge change.” Then I looked at the y-axes and noticed none of them start at zero. I hate it when people do that.

    1. I’ve heard a lot of people complain about not including zero in the y-axis, but there are plenty of cases where including zero would be a bad idea, and this is one of those cases.

      The problem with starting the y-axis at zero for the current discussion is that the variance over the last 10 years (~5k), would be reduced to a few pixels on the graph! So that really wouldn’t be helpful at all and I’d be asking Andrew to redraw the graph so we could see what he’s talking about.

      The mean VMT in the last 10 years appears to be about 240k and the variance around 5k. That tells you how to set the axis, and a quick glance at the y-axis tells you how far away from zero we are in this time period.

      However, had we been talking about the rise and fall of horses mile traveled (HMT) compared to vehicle miles travels (VMT), then we’d need graphs going back 100 years or more. In that cases, the mean VMT would be around 100k with variance around 50k or whatever. Then you’d be right, we should probably include zero. (Although, even that would probably swamp the variance of HMT).

      1. I didn’t draw the graphs, the St. Louis Fed did. You are welcome to ask them to make you new ones.

      2. The variance should only be a few pixels on the graph, because it’s just not that big! The total VMT in 2011 is 97% of what it was at the high point in 2007. Starting the axis at zero would allow you to see that there was a slight decrease, but only that.

        When you start the axis at over 90% of the all-time high, your visual system misleads you as to how large that small decrease actually was.

      3. Your visual system misleads you as to how large that small decrease actually was.

        It seems like this a problem only a simpleton would have. I’m pretty sure I learned to read graphs in about the sixth grade.

      4. @Eric. No, the variance over the 10 years shouldn’t be a few pixels. This is barely even a debatable point. This whole discussion is about how VMT has decreased. If you draw a graph where the decrease is no longer visible, then it’s not a useful visual to illustrate the point.

    2. Then I looked at the y-axes and noticed none of them start at zero. I hate it when people do that.

      You understand, of course, that there are reasons people draw graphs this way??

      1. Yes. It’s because they want to make a relatively small change look larger than it really is. The fact that there’s a decrease at all is important, but it’s still a relatively small decrease and it would be nice if the graph allowed you to see the magnitude of the decrease for what it really is.

      2. You’ve clearly misread the graphs. The drop is nearly 10% per capita!

        Also, the graphs do that because it’s easier to understand what’s happening. The axes are clearly labeled, and are very easily understood. No one is attempting to mislead anyone, and I take massive offense at you insinuating that I am, in fact, doing that. This graph lets you see things for what they actually are.

        Furthermore, imagine how ridiculous many graphs would be if all axes had to start at zero:

        1) You could never have negative numbers
        2) Displaying values with very low absolute bounds but rather high typical bounds would become impossible. Imagine, for example, temperature graphs starting at absolute zero! What an incredibly stupid idea. Should I have started the x-axis of my graph at the big bang as well!? Or is zero AD good enough for you?
        3) Sometimes you are interested in the variance and not the absolute number, these graphs would be useless, because you couldn’t see the variance. For example, you care how stock prices , commodity prices, economic statistics, etc. change but their absolute value is not as interesting. Oh my god! This is exactly the case we’re in here! Who would have thought I would have tried to pick the tool for the job?

        I am trying to illustrate a trend. That’s the point, if you have a problem with that, fine. But it certainly is your problem. There’s nothing wrong with these charts.

      3. 1) You could never have negative numbers

        I should have said the value axis should include zero, when both positive and negative numbers are a possibility. Since it’s not possible to travel a negative number of miles in a vehicle, that’s sort of irrelevant for the first and third graphs.

        For the second graph, the fact that zero is included does enhance the readability of the graph. You can see by looking at the graph that the growth in 2004 was about twice as much as the growth in 2003, because the data point for 2004 is about twice as far away from zero as the data point for 2003.

        Suppose you were only looking at the years 2002 to 2006, so you decided to chop off the bottom of the graph at around 2,000 units. The 2004 data point would be 4x farther away from the bottom of the graph than the 2003 data point. The only way to make a comparison would be to pull out a calculator and divide the numbers (or divide the numbers in your head). But if you have to do arithmetic to make a valid comparison between two data points, I would argue that the graph is only marginally more useful than the data table from which it was derived.

        2) Displaying values with very low absolute bounds but rather high typical bounds would become impossible. Imagine, for example, temperature graphs starting at absolute zero! What an incredibly stupid idea. Should I have started the x-axis of my graph at the big bang as well!? Or is zero AD good enough for you?

        The x-axis and the value axis are completely different entities. Of course it’s valid to limit the x-axis to the area you want to examine. But when you remove zero from the value axis, the utility of the graph goes way down. All you can get out of the top and bottom VMT graphs (without pulling out a calculator) is that there was an increase until 2005-2007, and then there was a decrease. If that’s all you want to communicate, fine. But if you want to show people how much the VMT decreased, including zero in the axis lets you see it, no calculator (or mental arithmetic) required.

        3) Sometimes you are interested in the variance and not the absolute number, these graphs would be useless, because you couldn’t see the variance. For example, you care how stock prices , commodity prices, economic statistics, etc. change but their absolute value is not as interesting. Oh my god! This is exactly the case we’re in here! Who would have thought I would have tried to pick the tool for the job?

        Stock prices change relatively little per day (a change more than 5% in either direction is rare), so it has become customary to show a tightened graph. Even there, you still should care about the absolute change. Suppose two stocks each fell by $20 today, and you presented this information in side-by-side graphs where the top and bottom of each y-axis varied by about $25. If the first stock is Berkshire Hathaway (which trades at about $130,000/share), a $20 drop is practically meaningless. If the second stock started around $30, a $20 drop might well indicate the company announced something catastrophic that is likely to lead to bankruptcy. Two graphs that look identical have very different interpretations, precisely because you’re narrowed in on the min/max values and are ignoring the big picture.

        Now, I doubt you had any intent to mislead people by presenting these graphs. I apologize for implying that you did. I still do maintain that the first and third graphs are misleading though. The decrease in total VMT from the high point to 2011 is on the order of 3%, but it’s just presented so much lower on the graph. Someone who just looks at the graph without pulling out a calculator will assume the total VMT was halved. Clearly that hasn’t happened, and anyone who thinks about it for a minute will realize that can’t possibly be the case, but a massive change is still what you see on first glance. Why not include zero on the y-axis, and make the actual change clear from the start?

      4. I don’t think there is a way to coax that FRED tool into getting a graph with a graph starting a zero, since their data goes to 1970, and VMT wasn’t zero in 1970.

  4. It would be neat to see this data at a local level and be able to look at economic growth, transit & vanpool ridership, bicycle miles traveled, etc. over the same period. I guess we should throw in total operating expenses for public transit agencies too.

    I really wish we were spending a bunch of money to bore a subway tunnel for LINK to Ballard rather than the HWY 99 tunnel.

  5. It looks as though you used a sum of the annual VMT divided by the sum of the Civilian Non-Institutional Population. The problem is that it’s summing the monthly population as well as the VMT, so your units are out of whack. What you want is a sum of VMT divided by the population at the end of the period. I can’t figure out how to make the FRED graph do that though.

    The trend line is the same though, so I suppose it doesn’t matter.

    1. Yes, the units in the bottom graph are sort of nonsensical, which is why I put it at the bottom. It’s still interesting to see.

  6. I’m not the graph wizard, but perhaps it would be nice to see the above graphs side by side with graphs depicting some values for age of the population. I think that the aging of the population is also playing a role here.

  7. Gas prices climbed higher than normal in the summer of 2005 and then when Katrina hit, they remained high through the winter, which was a historical aberration. A prolonged period of high prices does a lot more to change behavior than temporary price shocks. Other than immediately after the crash, fuel prices have never retreated back below where they were in 2005 and the public has come to expect a gallon will be more than $3. Some people have responded by buying more efficient vehicles, but for many that’s not an option. They have to change travel patterns – Combining more trips, avoiding trips by shopping online, switching to transit or other modes for some trips, moving closer to work, etc.
    Also, because of the high price of fuel, higher interest rates, and other rising commodity prices – all inputs to construction (materials costs rose due to the massive amount of rebuilding going on in 05-06, plus Asian demand) the building industry started to fall apart in 2006 prior to it eventually bringing down the whole show a year and a half later. Construction requires a lot of vehicle travel, and without millions of workers travelling to job sites, well…
    There are certainly other stories here too, infill, expanded transit, revival of central cities, generational preferences… But those are my bets for the 2 biggest drivers for lower VMT in the past 10 years. Anyone have studies to tease out causation? That would be a good topic for a brilliant grad student…

    1. There is an interesting conversation on that topic link I provided, it has some data, but is partly anecdotal. It does seem that young people, regardless of income, are driving less across the board than previous generations have.

    2. The odd thing is, when gas prices hit $4 in 2008 people thought the world was about to end and they’d never be able to fill their gas tanks. Then this year it was $4 again for several months and people barely mentioned it. Even though the economy is worse now than it was then.

      1. People get “adjusted” to the price of gas. The first time it spikes, is shock and horror. The next time, is surprise, the third time, its business as usual. The gasoline industry has figured this out. I have to wonder if we did charge a higher gas tax, or sales tax on gas purchases, would the industriy compensate by lowering the price of their product to keep the prices similar to what we have to day which seems to lead to stronger sales? When prices hit $4.50/gal it was noticed at the pump so they need to find that soft spot where they can still make gross profits and keep the product moving (and the economy).

  8. Maybe this has something to do with it:!ctype=l&strail=false&bcs=d&nselm=h&met_y=Autos_driver&scale_y=lin&ind_y=false&rdim=state&ifdim=state&tdim=true&hl=en_US&dl=en_US&ind=false

    Declining # of autos per person since 1990. Car-sharing studies have found that people who don’t own a car drive much less, even when a car is available. So less cars in HH’s may have finally started to impact travel. I’d still go with price & employment as the main factors. The trend so far isn’t significantly greater than the downswing in the 30’s, so until we have a roaring economy, predicting an end to VMT growth is premature.

Comments are closed.