Downward variation

I came across two strange articles recently. One claimed reducing speed limits reduced the speed of traffic. Another claimed reducing speed limits reduced crashes.

Of course it’s not that simple. Reducing the speed limit in areas of heavy fog may have reduced speed more than the fog alone did. Reducing speed limits during snow storms demonstrated a classic case of regression to the mean.

Drivers exiting Virginia on I-77 southbound frequently head downhill into fog. About 10% of crashes are in fog but they account for close to half of all vehicles. That’s because chain-reaction crashes are more likely in poor visibility.

As on most American highways, the speed limit on I-77 is too low. It’s 65 when it should be at least 70. In clear weather.

When fog rolls in the formula says 70 is too fast. Unfortunately it’s not a very good formula, the stopping sight distance equation. What is says is if you aren’t paying attention to driving and you aren’t willing to slam on the brakes, you would hit that jacknifed semi blocking all lanes ahead of you. But most people are paying attention and most people are willing to hit the brakes and most obstacles can be avoided by swerving instead of braking. So most people are going to drive faster than the formula says without getting in trouble. That’s why basing speed limits on design speed is a dumb idea.

The people setting the variable speed limits on I-77 tried to make them almost believable. They chose to set speed limits about midway between what people actually did and what the computer said they should do. Only half as unbelievable as an ordinary speed limit.

Based on limited data, apparently due to equipment failures, the Virginia Transportation Research Council said reduced speed limits definitely maybe work.

What didn’t work was reducing speeds far in advance of fog. It seemed like a good idea at the time, reduce the speed limit in stages rather than all at one. But that works about as well as reducing a speed limit to 40 because the highway ends half a mile ahead. It takes 500 feet to stop comfortably from highway speed, not 2,500. It takes about 200 feet to stop hard. We’re not going to spend a minute crawling at half speed.

The report observed a related delayed reaction: “This case study illustrated a common phenomenon observed during fog events where drivers must be exposed to fog for some time before decelerations occur.” You could argue that they should reduce speed immediately when they hit fog, but there is no need to slow them down a mile in advance.

One number to look for in this kind of report is the correlation coefficient, R2. If R2 = 1 your model perfectly matches the data. You may have determined a fundamental law of nature or run a bad experiment, but either way you can plot a perfect curve through all your points.

In real life experiments are messy. A coefficient around 0.9 is good. Think of the number as how much of the real world your model can predict. Virginia’s variable speed limit experiment has R2 = 0.32. Their model doesn’t predict much at all. Most of the variation in driver behavior is still unexplained.

Unfortunately Virginia’s experiment had a lot of missing data and it told drivers to slow down before they needed to. I would like to see a new experiment with more closely spaced speed limit signs, fewer false alarms, and better data collection. Because it’s certainly not impossible that a reduced speed limit combined with fog will be taken seriously, and maybe it could affect safety.

Ohio DOT recently claimed reducing speed limits during winter weather was effective in reducing crashes. Here is the reason I don’t believe them. According to WOIO, “ODOT put the plan in place following multi-car crashes on 90 in back-to-back winters, both of which included upwards of a hundred cars in each accident.” I bet if they raised the speed limit to 80 during snow we would have seen the same effect. It’s called “regression to the mean.” If a winter is worse than average you should expect the next year to be better, i.e. closer to average. If you do anything in the name of safety after a worse than usual year, you can expect an apparent safety improvement.

There’s a big underlying problem in most variable speed limit implementations. They don’t raise speed limits, only lower them. They’ll lower the fixed speed limit because a road is sometimes dangerous, and when the variable speed limit comes in they lower it even further when the road is actually dangerous. What if the weather stations are saying it’s a sunny day? Why can’t they raise the limit? Glenwood Canyon, Colorado is the only place I know where a variable speed limit is combined with a speed limit increase.

We should eliminate highway speed limits under normal conditions. When weather changes are imminent, set an 80 mph speed limit. When the road surface is wet shoulder to shoulder, drop the limit to 70. When there is standing water, sticking snow, or low visibility drop it further.

Some of us might start to believe those signs. But, to pick an example, when Massachusetts drops the speed limit to 40 on 100 miles of highway because it’s raining, we don’t believe that. Nor should we.

The opinions expressed in this post belong to the author and do not necessarily represent those of the National Motorists Association or the NMA Foundation. This content is for informational purposes and is not intended as legal advice. No representations are made regarding the accuracy of this post or the included links.

Not an NMA Member yet?

Join today and get these great benefits!

Leave a Comment

3 Responses to “Downward variation”

  1. Mikhael says:

    I’d take design speed over the current regime of arbitrary and capricious. At least design speed is tangentially, if overly conservatively, related to reality. Are you saying that design speed is dumb because it assumes a driver will see and react to an obstacle? What do you substitute for that assumption?

    • John Carr says:

      Design speed does not do a good job of determining actual traffic speed or safe traffic speed on high speed roads. There are a few problems with design speed. Vehicles have changed a lot over the past century. Obeying the recommended speed on a curve will prevent you from slipping on the bench seat of your model A. So what? Some agencies lump roadside barriers into design speed, but only when they need an excuse. The telephone poles next to the 55 mph road don’t matter, but a guardrail only 8 feet away instead of the ideal 30 will be used as an excuse to lower a speed limit.

      Maybe an agency will want to save a little money by making ramps 500 feet long instead of 600. To do that they have to lower the design speed of the entire road. They don’t have to change anything about the road, but if one ramp is one foot shorter than ideal they can choose to use that as an excuse to run speed traps everywhere.

      Design speed is a warranty: On a dark and stormy night, if you drive this fast you won’t spill your drink and you will stop in an emergency without triggering ABS.

  2. Hoggardz says:

    Speed limit really is the key to faster and safer roads. People have different capacity on speed and driving and it is hard to keep them at a certain level.