Map Monday: Comparing Safe Routes to Quick Routes in Driving to Work in the Twin Cities Metro

Here’s a map from a recent working paper co-authored by Mengying Cui and board alumnus David Levinson. The analysis, entitled “The Safest Path, analyzing the effects of crash costs on route choice and accessibility,” looks at Twin Cities road data and crash stats.

To make a long story short, crashes are expensive, and there are ways to avoid them that might take a little longer. That’s what we’re looking at here.


Here’s the key map:

Here’s a description of the map:

The differences of work trip flows between using the safest paths and the shortest travel time path are shown more clearly in Figure 2c, in which red lines refer to significant higher trip counts of using the shortest travel time path, while blue lines refer to significant higher trip counts of using the safest path.

Basically, as far as I understand it (and there’s a lot of scary math), the blue roads are ones people would take if they wanted to be safer, and the red ones are those that people would take if they wanted to be riskier-but-faster.

It looks to me like the older highways like 100, 169, and 62 do very poorly according to this analysis. Newer highways like 464/694 and 35E fare better. is a non-profit and is volunteer run. We rely on your support to keep the servers running. If you value what you read, please consider becoming a member.

, , ,

10 Responses to Map Monday: Comparing Safe Routes to Quick Routes in Driving to Work in the Twin Cities Metro

  1. Jeff September 19, 2017 at 9:58 am #

    What strikes me is there is an awful lot of red on the map, which I take to mean there are a lot of fast, risky freeways. The only bluish roads seem to be on the outer fringe. So fast roads plus higher traffic count equals risk. That makes sense.

    • Shawn September 19, 2017 at 10:38 am #

      I noticed that too. I bet there’s a pretty strong correlation between volume and accidents. I’ll read the paper to see how they quantified/qualified “safety”.

      • Shawn September 19, 2017 at 11:45 am #

        Okay, I Read the paper. They’re using crash rates * “cost” to define safety as the “cost per segment” of an accident. This provides frequency and scale, weighting fatal accidents more heavily and fender benders less so. The models also take into account volume, speed, and segment length – and a final parameter for urban and rural.

        “The work trip flows based on the safest path, however, do not coincide with flows based on the shortest travel time path, which reflects a significant difference between those two types of path. For instance, travelers need much more time to travel following the safest path, and generate greater crash costs following the shortest path. Mitigating such a conflict is an efficient way to improve the performance of the network.”

        They didn’t present a standard ANOVA to isolate which variables have greater impact on the results, but from what I can read of their analysis…. Yes, speed and density are the biggest contributors to accident “cost”.

        The other thing I’d like to see is an analysis of what would happen if 20% or even 50% of people chose “safe” routes over “fast” routes. I have a feeling that would change the shape of their transit maps greatly.

        FYI, Meng is a PhD student in Civil Engineering and David is a well titled instructor in that department: “Chair of Transportation Engineering”, and “Director of Network, Economics, and Urban Systems Research Group”.

        • Dan September 19, 2017 at 12:02 pm #

          In their regression results table, you can see the coefficients of traffic volume (V), segement length (L), speed (S), variance in speed (S_var), and whether or not the road is urban (U). The results are that volume has a much stronger correlation than speed, which is on the edge of significance. Variance in speed (i.e. does the road move fast sometimes and slow other times) is more significant than speed.

          I would say that volume and length are the two biggest contributers to cost…although with pesudo R^2 values uniformly less than 0.1, it is hard to attribute much predicted power to their models.

        • Bill Lindeke
          Bill Lindeke September 20, 2017 at 10:40 am #

          Thanks for the clarification on the titles.

    • Dan September 19, 2017 at 11:33 am #

      That is not a correct interpretation of the data. If you click on the link, you will see in figure 1 of the report the crash cost estimation of the each road ‘link’ in graphical form. You will see on that map highways outline in red: they have the lowest crash cost per vehicle kilometer. Highways are generally known to be safer per vehicle kilometer than local roads, and this study bears that out.

      • Shawn September 19, 2017 at 11:47 am #

        Yea, but it’s hard to compare the density and lack of intersections of a highway to local roads. They’re apples n’ oranges, really.

  2. Lindsey Wallace
    Lindsey Wallace September 19, 2017 at 10:02 am #

    Now someone needs to do an analysis of safe routes vs. quick routes on a bike. It would be a useful investigation to respond to the criticism that bikers can just take side streets and the busier streets should be left for cars.

    • Shawn September 19, 2017 at 10:37 am #

      I want to see that study!

  3. Dan September 19, 2017 at 11:56 am #

    I have some problems with the study.

    1. What the graph attached to this post is showing is the difference between safest path and fastest path. The highways get much more traffic than other roads, so they show up in red or blue, the colors with high magnitude. Since the local roads are all in yellow, we cannot really tell if they are over- or under-used based on safety. A better metric would have been ratio between shortest and safest path, instead of difference.

    2. The magnitudes of traffic variance as shown are actually very small. The survey used MnDOT’s accident classification scheme, which values a fender bender at $7,600; a minor injury crash at $83,000 all the way up to $10 million for a fatal accident. The dataset for crashes is over 12 years. So when you see a red segment on the map, that indicates that over 12 years, that one link would have $30,000 more or less of crashes is people drove for speed or safety, respectively. However, the $30k only means 4 fender benders or 0.003 fatal accidents _per 12 years_.

    Now the segments themselves are from TomTom, and represent sections of the road between two places you can turn. TomTom’s data costs money, so I can’t tell exactly how many segments per section of road. For example, I don’t know if the section between the east and west off ramps of highway counts as its own section. For example, between the junction of 169 and 494 and 94 (northbound) I count 21 (or 35) segments on 169 and 13 (or 24) segments on 494. Since 169 is red on the map in the post and 494 is mostly blue, that means that if drivers switched to optimize safety rather than speed, they could prevent (21+13)*30,000/7600 = 134 fender benders over 12 years; or 2.2 incapacitating/fatal accidents. Roughly double both those numbers if the sections between off ramps also count.

    Conclusion: Ultimately, that isn’t that much of a difference. According to the paper, MnDOT reported 680 fatal or incapacitating accidents a year over 12 years in Minnesota, and those accidents represented 50% of the equivalent crash cost. If drivers switched from driving on 169 to 494 to optimize safety instead of drive time, this would reduce the number of fatal/incapacitation accidents by 0.1 per year. This is not a significant effect.

Note on Comments welcomes opinions from many perspectives. Please refrain from attacking or disparaging others in your comments. sees debate as a learning opportunity. Please share your perspective in a respectful manner. View our full comment policy to learn more.

Thanks for commenting on!