Electronic National Dialogue on Transportation Operations
Archive -- Read-Only

Re: Yellow Increased - Red Light Running Way Down

Posted By: Gene Quinn <equinnok@ix.netcom.com>
Date: Friday, 27 July 2001, at 3:52 p.m.

In Response To: Re: Yellow Increased - Red Light Running Way Down (Richard B. Easley)

I have read some of the posts in response to mine and appreciate the feedback. I have posted this as a response to Richard but it covers some points and questions raised by others too. I could not answer them all due to time constraints but I will provide additional comments as time permits.

The Virginia intersection where the yellow was increased is US50 at Fair Ridge Drive in Fairfax County, VA. The camera enforced approach is the westbound US50 direction. The volume on the approach is 37,000 [ADT 74,000 both ways]. The approach is downhill an estimated 3%. There are three through lanes and one each left and right exclusive turn lanes. The posted speed limit is 45 MPH. The 85th percentile speed is estimated at 50-55 MPH. There are large numbers of commercial vehicles and tractor trailers in the vehicle mix. This approach was identified by VDOT last summer as one of the worst in northern Virginia for entries on red, the condition presumably being the basis for the camera installation

What we do know from the data after the increase is that high voluntary compliance to the signal is not just possible but it happens with a yellow time of 5.50 seconds, a circumstance that was not the case with the four second yellow. The results from this increase to the yellow show that the 5.50 seconds results in a statistically insignificant amount of entries on red ((ie. high compliance) for the prevailing conditions. Richard correctly notes the drivers did not know of the increase. Go/No-Go decisions by drivers in response to yellow onset were the same just before the increase and just after as no time or reason for adapting was available to the drivers. There were no signs erected that said "LONGER YELLOW LIGHT AHEAD". The "after" observations and data, compared to the "before" data, show with certainty that large numbers of drivers were being protrayed by the four second yellow light as "running the red light". Why? Premature onset of red for the prevailing conditions existed with the four second yellow.

Richard expresses concern about the long held hypothesis about long term adaptive behavior, namely, that people driving will eventually figure out that the yellow is longer and that some drivers might start going when they otherwise would have stopped. Richard mentions "if the yellow phase was adequately designed the first time", presumably thinking it was. Form your own views now that you have more information about the intersection approach.

How does one distinguish between people entering on red intentionally versus those portrayed as such by premature red onset for the prevailing conditions? Start with an intersection with an unacceptable entry on red problem and apply the measure of effectiveness test for yellow by incrementally increasing the yellow and counting the results. First count existing entries on red with the existing yellow setting that may be causing the problem for some fixed number of cycles during a time of day when volume/speed conditions produce the greatest numbers (baseline). Increment the yellow up 0.50 seconds and count the same way the next day, assuming volume/speed conditions are the same. Increment the yellow again 0.50 seconds and count again on the third day. Repeat as necessary until entries on red approach zero. Usually no more than 3 or 4 increments are needed, sometimes less, to isolate optimal yellow. You may have to count over multiple days for lengthened yellows because absolute numbers of entries will drop way down. In other words, a greater sampling size is needed to count just a single entry on red event, as they become increasingly rare (fewer and farther between) with each increment. At the VA intersection noted above, I took fourteen hours of video (after the increase) looking to observe a RLR but could not. Doing this simple test reveals the intersection specific optimal yellow for the prevailing conditions. Since drivers do not know you are increasing the yellow during the test, drivers are not adapting. [The measure of effectiveness test works in reverse too (Not Recommended). I mention this simply to indicate there is a way to show how a shorter and inadequate yellow can easily portray increasing numbers of normal people as "running a red light" when they are not. Most newscasts regarding red light cameras that purport to show "red light running" are actually showing entries on red during the change interval produced by inadequate yellow time on the lights. If one starts with an intersection that does not have an entry on red problem, incrementally reducing the yellow will easily create one]

After you do the test, pick the yellow you want for the amount of entries on red you want and make that the setting on the light. Keep in mind that if you are truncating the yellow below that needed for achieving highest compliance, you are making a decision to portray some people in the future as "running the red light" when they very well might not be, although no doubt they will look that way to observers. Remember as you make the decision that, should there be an accident in the future with a truncated or inadequate yellow, you may have relegated an otherwise normal person driving normally to a difficult fate. Witnesses to accidents base their decisions about behavior of others on the visual reference frame provided by the lights, even when those lights are improperly timed or are inadequate for the prevailing conditions.

If entries on red at the VA intersection increase over time, we will certainly know with greater certainty that those future subsequent entries on red are indeed intentional (all other factors remaining constant). Knowing this should be a condition before extreme enforcement such as cameras is considered? Without doing the measure of effectiveness test as described, whether one's behavior is intentional or not in responding to the yellow onset is nothing more than an academic guess, regardless of how educated or well-intentioned it might be.

The kinematic formula has shortcomings. It is exclusionary and will produce entries on red by design, particularly if the parameters used are not wisely and realistically selected and kept in their proper context at all times. A simple example: Two signalized intersections exist and they have identical parameters that are typically used in the kinematic formula. However, the approach volume is twenty times greater and there are more lanes at one intersection than the other. The consequences of any inadequacies of the yellow from the kinematic formula at the low volume, single lane approach are much different than on a higher volume multilane approach. On the higher volume multilane approach, the absolute # of observations of entries on red because of inadequacies (even small ones of just a few tenths of a second) will skyrocket in the eyes of observers, especially during peak hours. Observers, including cross street traffic, pedestrians and bicyclists are unhappy with what they see. They understand intuitively that the entries on red they see are sources of increased potential conflict, are potentially very dangerous, and represent potential threats to their well being. More often than not, they are correct.

"The results of the first two comparative phases of this study indicate that driver behavior does not change significantly when faced with longer amber phases." ["Driver Response to Amber Phase of traffic Signals", Olson/Rothery, 1962]

"The most significant observation is that although increasing the amber duration changed the boundary locations between the regions, and therefore the percent of vehicles observed in each region, the behavior of traffic in accepting and rejecting the amber phase at various distances from the intersection remained essentially unchanged. It appears that traffic is unaffected by increasing amber duration either because the drivers are unaware of the increase or the drivers are aware but not affected by it." ["Clearance Interval at Traffic Signals", May, 1968]

"The results in table 3 show that the extension of yellow duration reduced the frequency of potential conflicts in all cases studied." ["The Influence of the Time Duration Of Yellow Traffic Signals On Driver Response", Stimpson/Zador/Tarnoff, ITE Journal, November 1980]

"It has frequently been claimed that if the yellow is "too long", more drivers will use part of the yellow as green. More drivers - it was argued - would cross after yellow onset with long than with short yellow" ["The Influence of the Time Duration Of Yellow Traffic Signals On Driver Response", Stimpson/Zador/Tarnoff, ITE Journal, November 1980]

"The data show that the percentage of last-to-cross vehicles clearing the intersection (T+0.2) seconds or more past the yellow onset was not appreciably changed by the extension of the yellow phase." ["The Influence of the Time Duration Of Yellow Traffic Signals On Driver Response", Stimpson/Zador/Tarnoff, ITE Journal, November 1980]

"Research has consistently shown that drivers do not, in fact, adapt to the length of the yellow." [Determining Vehicle Change Intervals - A Proposed Recommended Practice", ITE, 1985]

"Olson and Rothery reported in 1972 that their research showed that drivers were "virtually" certain to stop if their required deceleration rate was less than 8 feet per second squared and virtually certain to continue if the deceleration rate required was in excess of 12 feet per second squared" [Determining Vehicle Change Intervals - A Proposed Recommended Practice", ITE, 1985]

"The average implied deceleration rate of the group with the highest crash rate was slightly over 13 feet per second squared, and the deceleration rate for the group with the lowest crash rate was 8.5 feet per second squared" ["Effect of Clearance Interval Timing on Traffic Flow and Crashes at Signalized Intersections", Zador/Stein/Shapiro/Tarnoff, ITE Journal, November 1985]

"The exclusive function of the steady yellow interval shall be to warn traffic of an impending change in the right-of-way assignment." [Manual on Uniform Traffic Control Devices, (MUTCD), FHWA, 1988 Edition, pg 4B-15]

"Signal Operation Must Relate To Traffic Flow" [Manual on Uniform Traffic Control Devices, (MUTCD), FHWA, 1988 Edition, pg 4B-18]

"In reality, if the signal is warranted and is properly designed and operated, improved safety may result, but this is only a by-product of the main purpose for traffic signal control which is the "orderly movement" of traffic." ["Traffic Control Devices Handbook, FHWA, 1983, pg 4-3]

Entries on red caused by inadequate yellow time on a traffic light do not promote the orderly flow of traffic. Signal operation must relate to traffic flow, not the other way around. Who cares if a properly timed yellow light, which should produce the highest possible compliance with red onset by definition, results in some vehicles moving through the intersection on yellow as a result?

I asked a researcher a few months ago if there are any studies that confirm the hypothesis Richard is concerned about, that is, concern about long term adaptive behavior to a lengthened yellow? The answer was there are none. I can't find any myself and those studies I have located on the subject refute the contention, not support it. I would suggest that if such a contention were true, it would have been proven scientifically, not just once but many times. If there are no studies that scientifically validate a hypothesis as true, the hypothesis should not be used when establishing change interval timings. Speaking for myself, I could not justify arbitrarily truncating the yellow below the amount needed if no evidence exists to support the decision, especially if I am complaining about red light running on the other hand and cheerleading for red light camera enforcement.

Has anyone ever considered why, for the same driving population, the compliance at one signal can be almost 100% while the compliance with another signal can be miserable? Recent newspaper accounts indicate that entries on red are happening at the George Washington Memorial Parkway/Slaters Lane in Alexandria, VA at a rate of 40 or more per hour under the three second yellow. Contrast that figure with the 3-4 per day at US50 and Fair Ridge (5.50 seconds of yellow) over in Fairfax County,VA.

"Red Light Running" of course is not good. All of the data shows that. "Red light Running" accidents have grown three times faster than other accident categories since 92-94. There is a reason. To the extent possible, engineers should use the tools they have to minimize entries on red, lengthening the yellow being the simplest and least expensive way. Besides, it can be further adjusted later if needed. The degree of "red light running" at any intersection can easily be just as much a reflection of how well engineers are doing (or not doing) as it may be of how well people driving are doing (or not doing).

I see a respondent indicated that valid data is needed, perhaps inferring the data used in the release is not valid. For informational purposes, the release was based on official county red light camera enforcement records. Further, that same data indicates no statistically significant reductions in entries on red at another camera site after 8 months of enforcement. The yellow was incremented up there to 4.50 seconds from 4.00 seconds on 6/28/01 as part of an ongoing evaluation program by VDOT. Detailed violation and citation data from camera operations for all camera sites in Fairfax County was requested by VDOT but receipt of same from the camera vendor was still pending as of earlier this week.

Vienna, VA increased yellows times in MAR-APR, 2000 at both camera and non-camera sites. Other engineering improvements were made. The three cameras recorded less than one violation per camera per day in the last quarter APR-JUN01. No red "grace" is used in the camera enforcement operations. The engineering improvements, the most significant being the yellow increases, have made a world of difference.

If you have an intersection with an obvious entry on red problem, why not consider what I suggest and see what happens?

Hopefully, this post will add constructively to the important discussion about change interval timings and the contemporary entry on red problem. If anyone knows of a study (or studies) that confirms the validity of Richard's concerns about long term adaptive behavior, by all means post the information needed to find it.

Responses To This Message

Re: Yellow Increased - Red Light Running Way Down (views: 1253)
Joel Marcuson -- Monday, 30 July 2001, at 3:28 p.m.

Talking Operations Forum is maintained by Operations Dialogue Administrator with WebBBS 4.33.