It is amazing that the topic of setting adequate yellow light intervals for traffic signals continues to be a hotly-contested topic among traffic engineers. The yellow timing is an important variable for intersection design, and it is also important to drivers for another reason: Short yellows combined with photo enforcement is a proven method for generating volumes of red-light violations.
As has been shown many times throughout the years, a small increase of 0.5 to 1.0 second in the yellow duration can decrease the volume of camera-generated tickets by 50 percent or more. Cities such as Loma Linda, CA have documented red-light violation reductions of 80 to 90 percent virtually overnight. Contrary to claims by ticket camera supporters, those improvements did not disappear as drivers “got used to” the longer yellow cycles.
That makes it all the more interesting that the yellow interval timing debate rages on in the traffic engineering community. During one such recent discussion on the Institute of Transportation Engineers member forums, Jay Beeber posted a conversation he had with Karl H. Zimmerman some years ago. Zimmerman, along with co-author James A. Bonneson, published research on yellow light timing in 2003. Their “Effect of Yellow-Interval Timing on Red-Light-Violation Frequency at Urban Intersections” was subsequently published by the Transportation Research Board.
The Zimmerman-Beeber dialog that follows is quite insightful. While it probably won’t stop engineers from continuing to debate yellow light timing as they have for decades, it surely should give pause to those who argue for trimming yellows to bare minimums. The bolding and underlining was added for emphasis by Beeber.
Question: In general, we discussed the results of your research work with James Bonneson for the Texas [A&M] Transportation Institute. You indicated that your results showed that increasing the yellow duration beyond the value computed using the ITE Kinematic Formula provided additional safety benefits through a reduction in both violations and collisions. Could you confirm this?
Zimmerman Response: Confirmed. The ITE Kinematic Equation handles the basic physics of the situation well but does not specifically address other related factors such as signal conspicuity, traffic volume, flow rate at the time of yellow onset, and other moment-by-moment factors. One built-in assumption may or may not be valid: the 1.0 s reaction time (which, depending on the complexity of the situation, might not be adequate). The grade term is also important because it significantly affects the stopping distance of vehicles, and it should not be dropped, especially if the approach grade exceeds 3 percent.
Q: Specifically, I asked about Table 2-2. Effect of Selected Factors on Red-Light Violation Frequency, from report 4196 (aka Table 6-1 from report 4027). You indicated that the results for Yellow interval duration under the section of the table entitled “Effect of an Increase in the Factor Value” meant an increase over the value computed using the ITE Kinematic Formula. The results show a 53% decrease in violations after increasing the yellow duration by 1.0 second. Could you confirm this?
Zimmerman Response: Confirmed. That is what the value in the table represents. Keep in mind, this is an average reduction. Some locations might show a higher reduction, some a lower reduction. We did not do anything outside of 5.5 s yellow intervals, so we cannot say for sure, but it probably will work up to 6.0 s and maybe even beyond.
Q: We discussed Figure 2-2. Red-Light Violation Frequency as a Function of Yellow Interval Difference, from report 4196. You indicated that the point designated as 0.0 on the figure would be the value computed using the ITE Kinematic Formula. The data indicate that yellow durations beyond the value computed using the ITE Kinematic Formula, up to about an additional second, resulted in a reduction of red-light running violations.
Zimmerman Response: Yes, up to a 5.5 s yellow interval, as stated in the report. I see no reason why increasing yellow would not work for yellow intervals above 5.5 s if there was a violation problem at an intersection.
Q: We discussed Figure 2-8. Effect of a Change in Yellow Interval Duration on Crash Frequency, also from report 4196. The data indicate that yellow durations beyond the value computed using the ITE Kinematic Formula, up to about an additional second, resulted in a reduction of red-light running crashes.
Zimmerman Response: Confirmed-we saw this with field data collection in Project 0-4027.
Q: We discussed Figure 5-13. Effect of a Change in Yellow Interval Duration on Red-Light-Running from report 4027. The data indicate that yellow durations beyond the value computed using the ITE Kinematic Formula, up to about an additional second, resulted in a reduction of red-light running violations.
Zimmerman Response: Confirmed – again, we saw this with field data in 0-4027. Vehicle-by-vehicle approach data from another project going on at the same time (0-4022, which involved dilemma zone protection at high-speed signalized intersections) implied that this effect would be true even though we did not change yellow intervals in the latter project.
Q: We also discussed the issue of whether drivers adapt to the longer yellow durations. You indicated that, in your professional opinion, this does not happen to any significant extent.
Zimmerman Response: I do not believe adaptation/habituation is a major problem over long time periods. I also do not know of a recent, valid engineering study that confirms or refutes that belief, and the logistical, temporal, and financial challenges of performing such a study at most intersections would make such a study infeasible. One instance where such a study might be possible is at intersections equipped with violation cameras, eliminating the need for costly manual data reduction. (In effect, you have already done that study at multiple intersections, which shows just how simple it is if data collection and reduction are already done.) Instead, there is a lot of hand waving on the subject.
It is possible that the appearance of adaptation can occur if traffic volumes increase without changes to signal timing, especially flow at the end of green and especially in coordination situations. That would not be true adaptation but an artifact of the control scheme. It would make sense that drivers adapt to longer yellows and give themselves a second more of effective green time to get through the intersection. Still, I would bet the average driver couldn’t identify a four-second yellow from a five-second yellow, and might not even realize the yellow at one intersection is longer than at another intersection. Instead, providing a longer yellow interval may give drivers time to make “safe and prudent” driving decisions that all state driving statutes are founded upon.
I believe that the expectation of adaptation is a mental analogy to the actions of drivers when speed limits are changed. Except…the circumstances are very different for speed limits and yellow intervals. Drivers experience speed limits all day, every day, on every route upon which they drive. Yellow intervals are site and approach specific and may or may not affect a driver at any particular time. A driver’s speedometer is on the dashboard, in clear view, all the time, and speed is a continuous factor affecting driving behavior. A yellow indication is visible for a few seconds at most, on signal equipment that is visible for maybe 15 seconds at a time (or less). The two are not parallel phenomena, so driver reactions to changes cannot be expected to be the same. (There are some indications that adaptation to new, higher speed limits does not happen, either. Many of the “speeding” problems are actually due to artificially low statutory maximums that drivers recognize as artificial and then choose to ignore.)
[As a side note, using posted speed in the calculation of yellow interval duration will underestimate the resulting yellow time for prevailing traffic conditions, as implied above. The actual 85th percentile speed, or speed limit PLUS an appropriate correction (my best estimate: 10 mi/h) should be used instead.]
Q: We discussed concerns raised by the red light camera industry and their supporters with regard to increasing the yellow time by about one second over the value computed using the ITE Kinematic Formula. Specifically, the following claims have been made:
- If the yellow time is increased, drivers will adjust and run the red anyway.
Zimmerman Response: Not necessarily. The potential for violations depends on: traffic flow at the end of the green, cycle length (and the motorist’s perception of delay if he/she has to stop), uncertainty about what vehicles behind a subject’s vehicle is doing (e.g., “Is the truck behind me going to go? Because if it is, I’d better go through, too.”), platoon arrival and cohesion, oversaturation, coordination effects, weather, fear of enforcement (reasonable or not), and other transient effects at each yellow onset. (Others have found slightly different effects than this list, but that does not alter my point.) Adding more yellow time may relieve some of these conditions. Proper engineering of signal coordination will relieve others.
- If the yellow time is increased, drivers will expect a longer yellow time at intersections where it has not been increased and will run the red signal at those intersections.
Zimmerman Response: If this is true, then why do we (as engineers) use different yellow intervals for different speeds? Would those not create the same situation that is being described?
This assertion is illogical on its face. Drivers do not really know the length of the yellow interval, and I will assert that it does not really matter to them, either. Drivers need what they need when the event occurs, with minimal carryover to the next one (which could be minutes, hours, or even days later). Even drivers who drive the same route day in and day out may not have the same set of events occur each time, which makes learning specific ways to “beat” the system much more difficult-assuming that most of them even try.
- If the yellow time is increased, this will increase the opportunities for rear-end crashes due to an increased exposure of motorists to an indecision zone.
Zimmerman Response: The opposite should be true here: increasing yellow times should reduce the opportunities for rear-end collisions by allowing drivers to make “safe and prudent” driving decisions. In addition, there should be fewer errors in decision making either way-drivers who decide to proceed (rightly or wrongly) will have more time to reach the intersection and will be less likely to be red-light violators. Drivers who decide to stop will be able to stop less abruptly. Drivers who are undecided have a better chance of making a safe decision. And truck drivers (see below) will be less likely to be violators as well.
All three of these assertions go against the basic legal assumption that drivers will make “safe and prudent” operating decisions when driving. The law is written assuming that they will. These three assertions assume that drivers, in general, are incapable of handling any responsibility and will behave in dangerous and stupid ways when behind the wheel. A few drivers will, but the vast majority will not. The assertions focus on the (proportionally) very, very few who are not “safe and prudent” either through inexperience, fear, confusion, intoxication, illness, or anti-social behavior, and not the many, who are merely trying to get to work or home safely without more inconvenience than necessary.
Q: We also discussed how increasing the yellow duration might affect intersection capacity. You indicated that it was a mistaken belief that increasing the yellow duration by one second at the expense of the green interval would decrease capacity. In your professional opinion, increasing the yellow time in this manner does not decrease intersection capacity.
Zimmerman Response: For every green interval, there are two periods of lost time: at startup (typically about two seconds, although this amount varies by cycle and by traffic mix), and at shutdown at some point during yellow (which may be anywhere from two seconds to four seconds or more, depending on the circumstances-usually more under very low flows). For operational analysis, both startup and shutdown lost time are combined and put at the end of the green interval. Then, for simplicity’s sake, the total lost time for the green interval is assumed to be equivalent to the yellow change interval (Y) plus the red clearance interval (AR). This is “easy,” but it can lead to the assumption that Y+AR = lost time. As a result, to “minimize” lost time, Y+AR should be as short as possible.
The reality is that lost time is not truly dependent on Y+AR, may be more or less, and does not change much when yellow interval duration is changed. This situation is actually a trade-off between safety and operational efficiency – you can get an extra second or two of “efficient flow” by trimming a second or two off the yellow interval, but you will also get a significantly larger number of red-light violators because drivers are not getting the warning they need to prepare to stop. Field data show this.