Can You Build Your Way Out of Congestion? Let’s Look at the Data

“You can’t build your way out of congestion”. Although this is a longstanding liberal dogma, I’ve never really subscribed to it, at least in the theoretical sense that is seems to be brought up. But recently I noted in comments to an article that Kansas City seems to move pretty well. Recently I found a good source of data from the Texas A&M Transportation Institute that sheds some light on this question.

There’s all kinds of data going back to 1982, so I had to select data to chart. Choosing freeway lane miles per capita was an obvious choice for how well built an area’s freeways are. For congestion, I chose annual hours of delay per auto commuter, because how long you’re stuck in traffic is the most noticeable and obnoxious thing for motorists, not the value of your time, the congestion index, or other more abstract metrics. Here’s the nationwide average data.


Now here’s the data for the Twin Cities. Although the scale is exaggerated, the slight dip in freeway miles per capita in the early 1990s is still a deep chasm. Basically nothing substantial got built through the 1990s while population continued to grow. Finally a series of funding bumps leveled off the decline, and eventually we are back at the 1980s level.


In both cases the congestion curve actually bears more relation to VMT (Vehicle Miles Travelled) than lane miles per capita. Congestion delay was reduced because the Great Recession cut the size of the economy. The well-documented aversion of millenials to driving may also be a factor, since VMT have been slow to recover since the Recession, as shown in the next chart.

Regardless of recent VMT trends, traffic projections assume congestion will worsen. Even the St. Croix Crossing, which has been made out by the anti-car forces as a paragon of excess, is forecast to be congested in the next 50 years. (And it’s worth noting that with the signalized intersection that Oak Park Heights insisted on, the capacity of the Minnesota approach road is a lot less than the traffic apocalypse that was US 169 in Bloomington).


But let’s not abandon all hope of proving that freeway lane miles can make a difference in congestion. Everyone knows that Kansas City flows better than New York City, so another comparison would be how congested various cities are compared to how much they’ve built to date. Obviously there are cultural and geographic differences influencing annual hours of delay per auto commuter, but if there is a trend it could be revealing. Thus I’ve run the data through a data smoother to see if there’s a trend that’s getting masked by some of the choppiness.


A couple of things are apparent, at least with the raw data it looks like there is a noticeable decline in congestion as lane miles go up sharply . It also looks like there may be a noticeable increase at the end, once lane miles fall below a certain point. The smoothed data suggests that on average, building more lane miles really does make a difference in congestion.

As for the outliers in the middle of  lane miles, I’d suggest that Miami, Phoenix, and Tampa have a large number of retirees that don’t need to drive in rush hour. Washington DC has a large number of government office workers,  San Francisco has significant geographical barriers, LA, Houston, and Atlanta have significant car cultures. The Twin Cities is a low outlier, I’d suggest despite a lot of choke points at under-built river bridges, overall our unique “grid” system to freeways makes it easy to route around congestion.

I also looked at the effect when arterials are factored in (using a conversion factor) from a different source that did a . Rather than see the same pattern, to my surprise this is what I saw.


I see no obvious correlation. This data is from back in 1999, but if you take it at face value it seems contrary to conventional wisdom that a good arterial system can relieve freeways. I’d suggest that there’s a strong motorist bias towards freeways, and directing resources other directions can even be counterproductive if there is an opportunity cost of not improving the freeways directly. Wide suburban-style roads are nice to get quickly from your neighborhood to the freeway, but it seems the quantity of them has little effect on the overall congestion on the area. In real life, witness what a horrible disaster area I-494 in Bloomington is at rush hour, compared with how well American and 77th move, both wide suburban style roads designed specifically to attract traffic from I-494.

As a side note there was an article recently that suggested that transit, while obviously worthwhile, should not be promoted as reducing congestion. So what is the effect of transit on average hours of delay per auto commuter?


Looks like the most congested cities and the ones with the best rail networks are the one where transit makes the biggest difference to motorists. In other cities apparently most people that can drive already do so. Still, it may be worth promoting it as such as kind of a white lie. As part of society I agree that we should have transit, but there’s a lot of people around me that are either apathetic or against it so promoting it as something to get other people off the roads they drive on may be the only way to get broad support. Like this Onion article.

As a final side note, here’s freeway lanes miles over time for some of the cities. In addition to Kansas City, Minneapolis, and the national average, I chose Chicago because of its famous congestion, Phoenix because of its aggressive highway expansion due to a local sales tax, Houston because of its famous auto and suburban growth culture, Portland because it’s the anti-Houston,


Overall, my views remain unchanged. It’s always possible to build our way out of congestion in a theoretical sense. The street in front of my house has been “built out of congestion”; with only two lanes, it’s not filled with people induced to come down from Blaine and drive back and forth on it just because the capacity is there.  Maybe we could build I-35W out of congestion with 15 lanes in each direction. Of course we can’t as a practical matter, but it’s still worthwhile doing something; if we can’t fix I-35W we can make a difference elsewhere rather than just repeating the dogma and doing nothing.

And of course self-driving cars are on the way, which will be the ultimate solution to the problem. Even if there’s still a bit of congestion, taking a nap or working on your laptop will make it not wasted time.


About Monte Castleman

Monte is a long time "roadgeek" who lives in Bloomington. He's interested in all aspects of roads and design, but particularly traffic signals, major bridges, and lighting. He works as an insurance adjuster, and likes to collect maps and traffic signals, travel, recreational bicycling, and visiting amusement parks.

11 thoughts on “Can You Build Your Way Out of Congestion? Let’s Look at the Data

  1. Wayne

    So now accepted truth with plenty of data and research behind it is “liberal dogma?”

    If your theoretical solution is replacing everyone’s houses with new highway lanes it has no place and no basis in reality.

  2. Alex CecchiniAlex Cecchini

    I’m not really sure why you used a smoother on a non-time series dataset. The order of the cities is mostly random but for the descending nature of one data point (freeway lane miles per capita), and therefore smoothing intentionally masks the (very important!) variation)

    A better way to look at this relationship is like this:

    I’m not sure what cutoff you used for metro areas, so I selected ‘large’ and “very large” metros from TTI’s database. There is a slight negative correlation (r^2 = 0.1) between freeway lane miles and average delay. But, if one were to try to take this at face value, we would need to build 2,757 freeway lane miles to save the average commuter 15.7 hours per year. At $20m per lane mile (perhaps conservative for urban areas), that comes to about $55 billion in costs (excluding annual maintenance/operating costs & the cost of upgrading all the collectors/arterials to feed said freeways). Simple math using current number of commuters & time costs from TTI says the MSP metro area would save $385 million a year for commuters. Or, $19.2 billion over the next 50 years, well below the $savings (which aren’t actually realized but for freight passing through the metro or other workers actually on the clock while driving).

    It’s obviously much more complicated than that. There may be other benefits (although using the same data since 1982 I found no correlation between lane mile growth & regional GDP growth: ). But the whole point many make here is that there are many other costs we never consider when we build or expand a freeway, arterial, etc. Safety, pollution, climate change, land opportunity costs, etc.Things that biking and transit do a much better job mitigating, even if they are a tad more expensive for the public sector per user (transit) or have longer travel times.

    1. Alex CecchiniAlex Cecchini

      Just another follow-up, TTI’s annual delay due to congestion is based on some bad assumptions (see here: for discussion on free-flow speed assumptions and discounting those not in cars for commutes, about 14% of the MSP metro).

      But even then, delay itself is a bad predictor of the thing that really matters, accessibility. Using the UMN Accessibility data I plotted weighted average job accessibility for drivers in a 40 minute window against metro-wide average commuter delay hours:

      Metro areas with more delay tend to have better job (and one could assume amenities like shopping or eating) accessibility. People are willing to make tradeoffs – a few extra hours per year in the car for the chance at more jobs (or, other things not counted at all in an analysis that only looks at travel time in a car). We should recognize that since people are willing to make all sorts of tradeoffs, including commute time, that price signals would be a more effective way to reduce congestion on freeways than simply building our way out.

    2. Wayne

      An r-square of .1 is meaningless. You literally cannot make any meaningful claim at that point. Which I guess was your point, but it bears repeating.

  3. mplsjaromir

    I find the chart labeled “Annual Hours of Delay With and Without Transit Service ” most intriguing. There is a correlation between income levels and annual hours of congestion.

    The greater the income level for a metro area the more likely it will see an increased level of traffic delay. The major outliers seem to be Houston and LA, who rank 58th and 43rd respectively in terms of income but yet have high levels of delay. At least LA has great weather.

    The other outlier is Minneapolis-St. Paul, this metro ranks 5th overall for average income yet appears to rank low on delays.

    From what I can surmise, people don’t really care about having low levels of traffic congestion. Kansas City may have built its way out of congestion, but the region is only growing at 2% annually. Well below MSP’s and the nation’s average for growth. For most traffic congestion levels do not factor into their decisions on where to live.

    1. Wayne

      I’d be interested to see population growth vs. congestion, as I imagine that would have more of a relationship than lane-miles per capita.

      1. Alex CecchiniAlex Cecchini

        Couple different ways to look at it:

        The bottom one is fairly telling to me. Basically, the ratio of a metro’s population growth rate to its freeway lane mile growth rate had no correlation to the % change in annual motorist commuter delay. In other words (according to this data), if population growth outstripped freeway building, congestion may have gone up but comparing across 100 metros finds that those who kept up better on freeways did no better.

        Of course, this is a snapshot of two discrete points in time (1982 & 2011), AND it masks all the other social/economic forces at play within metros. Make conclusions at your own risk.

    2. Wayne

      Also, woah — I had no idea MSP had a higher median income than the Boston MSA. What year are these results from?

  4. Monte Castleman Post author

    I figured someone with better knowledge of statistics and graphs would chime in. I flunked out of civil engineering, but as a pro-motorist voice here I figured it was up to me to at least try. I used smoothed data because I didn’t know any better way to try to spot trends, the cutoff was 1.5 million people to be sure to include Kansas City (which I’ve personally witnessed how well the traffic seems to flow and). An alternative title was “What’s (Not) the Matter WIth Kansas).

  5. David Greene

    Anti-car? The St. Croix crossing is a collossal waste of money that should offend all of us. Call a spade a spade.

Comments are closed.