By: Richard W. Lyles, Ph.D., P.E., William C. Taylor, Ph.D., P.E., Tracie Leix, P.E., M. Abrar Siddiqui, Ph.D., Bilal Z. Malik, Gregory Siviy and Tyler Haan; Contributing Authors
Priorities for the design of roadways have shifted over the years—from a primary emphasis on increasing capacity to considering the purpose of streets and roads in the context of specific settings (often referred to as context-sensitive design).
A technique that has gained popularity in recent years is the “road diet” where traditionally designed four-lane roads with two lanes in each direction have been converted/reduced to three lanes (often with the addition of bike lanes) with one lane in each direction and a center left-turn-only lane. Such conversions have potential impacts on both travel operations (delay) and safety.
A study was recently done in Michigan where 24 road-diet projects were examined with a focus on travel delay and safety for typical four- to three-lane conversions. The sites were initially identified by field engineers familiar with the site. To be included in the study, sites needed to have three-year before-and-after crash data available; consequently, road diets completed in the last couple of years were not included. The objectives of the study were:
Determine the safety-related impacts of the conversions;
Determine the delay-related impacts of the conversions;
Develop a statistically sound crash modification factor for conversions; and
Develop a guideline that addresses/incorporates the results that would be of assistance to various agencies in deciding when it might be desirable to implement such road diets.
The full report is available online at: www.michigan.gov/documents/mdot/MDOT_Research_Report_RC1555_376149_7.pdf and includes a literature and state-of-practice review, anecdotal observations regarding pedestrian and bicyclist use of road-diet streets and comments regarding the successful implementation of road diets in communities. This article is focused on the operational and safety aspects of road-diet implementation.
One of the most compelling findings of the study was that there is tremendous variation in the results of road-diet implementations. While most sites result in a reduction in crash frequency, for a few there were crash-frequency increases. While road diets are effective in decreasing left-turn crashes by taking the turning movement from the through lanes and putting it in a two-way left-turn lane, right-lane crashes may increase because of the resulting traffic volume increase in that lane. So, whether there were increases or decreases would appear to largely depend on existing conditions at the implementation site.
Give proper credit
While there is a fair amount of experience and some formal literature on the effects of road diets, much of the information is anecdotal rather than the result of formal studies. Notwithstanding the informal nature of the information, as a general statement, crash reductions are experienced with most installations. In the jargon of safety analysis, these reductions can be expressed as crash-modification factors (CMFs), where a number less than one indicates that crashes have been reduced and a number greater than one indicates that crashes have increased. There is, however, significant variation in the magnitude of the reduction. While there appears to be significant “natural variation” in crash-reduction percentage, variation also is introduced because of the differences in the before/after geometry and operating conditions.
Most studies did not result in estimates of CMFs per se. The most reliable estimates of CMFs (for many different safety treatments) are generally thought to be found on the FHWA-funded website Crash Modification Factors Clearinghouse maintained by the University of North Carolina Highway Safety Research Center (www.-cmfclearinghouse.org/). From that source, road diets are estimated to result in the following CMFs for unspecified roadway types in urban areas: all crash types/all severity levels = 0.63; all crash types/all injury crashes = 1.0; all crash types/PDOs only = 0.54; angle crashes/all severity levels = 0.63-0.76; and rear-end crashes/all severity levels = 0.59.
To accurately determine the impact of road diets, there should be an assessment of what is happening in the area even without the road diet. For example, if an area is already experiencing a general year-to-year decrease in motor vehicle crashes, the effect of the road diet (or any other significant change in the street and road system) could be overestimated—that is, a road diet might be credited with crash reductions that might have occurred anyway.
If traffic volumes are decreasing, fewer crashes could be expected in general. The effects of road diets (or any road/street system change) must be adjusted (or controlled) for the background variation in crashes. In most of the studies reviewed, there was not sufficient control for that background variation in crash trends.
While this study was focused on the safety and operations impacts of road diets from a motor vehicle perspective, numerous sites were visited and anecdotal observations made. While not all sites were explicitly addressed to pedestrian/bicyclist issues, observations included that provisions for pedestrians and bicyclists are important when there are existing pedestrian/bicyclist generators on the site and/or when the road diet is part of a larger plan for an area. If pedestrian/bicyclist provisions are included in the road-diet area, they need to be clearly and consistently marked. While there appears to be a need for additional information/education regarding appropriate use of the road and pedestrian/bicyclist facilities at road-diet sites, supplemental signs indicating crosswalks and bike lanes should be considered for routine inclusion at road-diet sites.
Careful of the traffic
From an operations (level-of-service) perspective, the argument against road diets is that since capacity is reduced through lane reductions, this may result in increased delay. The objective for the operational analysis was to determine if there was a common traffic volume at which a four- to three-lane conversion will be likely to fail from a traffic-operations perspective. Failure was defined for the study as the level of service dropping to D or below.
The rule of thumb in wide circulation is that a road diet should not be implemented when average daily traffic (ADT) exceeds 20,000 vehicles per day. Since the peak hour is not necessarily always the same fraction of ADT (e.g., in a tourist-oriented area, there may be a more uniform flow over the course of a day), the approach taken for this project was to look at the volume that leads to failure over the course of an hour. Examination of the limits of the traffic volume for which a road diet is appropriate is important, since increased delay is a concern that typically comes up when road diets are proposed.
Several typical sites were evaluated using standard highway capacity software (i.e., Synchro). The operational analysis of the several sites provide reasonably consistent results and support a guideline that suggests that four- to three-lane road-diet conversions can result in significant increases in delay for ADTs over 10,000, depending on the percentage of traffic experienced during the peak hour.
Much more importantly, four- to three-lane road-diet conversions can lead to increased delay when peak hour volumes exceed 1,000 vehicles per hour (vph). So, while a general ADT guideline can be used, it is far more important to look at peak-hour volumes. It has been argued that the suggested guideline of 1,000 vph is too low. However, if a general lane capacity can be assumed to be about 1,800 vehicles per hour, for a site with signalized intersections the 1,000-vehicle threshold makes sense—if signals reduce capacity, a nominal split of 55/45 suggests that 1,000 vehicles in the primary direction would be close to capacity.
However, it is clear that local conditions such as varying geometry, significant variation in turning movements and variations in cross-street traffic can have a significant impact on the viability of any proposed road diet. Thus, while an initial culling of potential road-diet sites can be accomplished using the general guidelines above, in all instances a detailed operational analysis of the corridor, including operations at each intersection, for both four- and three-lane sections should be undertaken before the road-diet conversion is seriously considered. Such analyses are straightforward and inexpensive.
Slimming rates
The crash analysis and development of a dependable and defensible CMF was a key part of the study. This was complicated by the fact that there was considerable site-to-site variation in crash frequencies. However, in almost all instances, there was a reduction in the number of crashes. Examination of the background trends (e.g., the city-wide trend over the several years of the analysis period) showed that in all cases there was a trend toward lower crash frequencies over time. That is, while the road-diet implementation sites generally showed reductions in crashes over time, there was already a trend in crash reduction at all sites. Thus, it was important to consider this background trend when assessing the effects of the road diets.
The most appropriate methods for controlling for background trends were a simple control for citywide trends and the consideration of comparison sites. The latter requires the identification of comparison sites that are virtually identical to the road-diet sites. These can be used to identify background trends and as examples of what would have happened at the road-diet sites if the road diet had not been implemented. Good/acceptable comparison sites could be identified for only a few of the 24 sites and none of the eventual comparisons gave statistically significant results. What that means is that the calculated CMFs for specific sites were not statistically different from 1.0. A CMF of 1.0 means that the crash frequency does not change between the before-and-after analysis period as a result of implementing some safety measure.
Notwithstanding the difficulty with comparison sites, the average CMFs, adjusted for background citywide trends, were calculated across all 24 sites. The result was that the overall (simple) CMF was estimated as 0.63. The 0.63 figure is the CMF without considering any citywide background trend. However, the CMF is a much more modest 0.91 after adjusting for the background trend. Considering only those crash types expected to be affected by the road diet (not necessarily only reduced), the adjusted CMF was 0.90. Considering only those crash types expected to be reduced by a road diet (“correctable” crashes), the adjusted CMF was 0.59. Use of the latter is problematic since there are typically offsetting changes in crash-type frequencies. Only the CMF for the correctable crashes was statistically different from 1.0. What these numbers show, on average, for the Michigan sites is that there are crash reductions for crashes that road diets would be expected to decrease (for example, mid-block left-turn crashes). However, these reductions are typically offset when all other types of crashes are considered and the overall reduction is adjusted for existing background trends. Thus, the more modest CMF of 0.91 is most realistic. While the best estimate of a usable CMF is 0.91, it should be noted that this is not statistically different from 1.0 and is an average across all sites. Perhaps more importantly, there is a great deal of variation from site to site.
Changes in crash severity due to road diets were examined and the distributional shift over all sites was estimated (and then compared to statewide changes). The finding was that although there was a slightly more substantial shift to less severe crashes for the road-diet sites, it did not seem operationally significant. Moreover, the shift could have easily been due to changes in operating speeds or enforcement rather than the road diets themselves.
Results may vary
As a consequence of this study, it is apparent that road diets should not be oversold with respect to expected benefits, especially safety benefits. Actual benefits of a road diet can vary significantly by site. Both the pros and cons associated with road diets need to be presented and thoughtfully discussed with the community. Use of social media to discuss (and resolve) road-diet proposals can be successful as an adjunct to traditional public hearings and other traditional community-involvement techniques.
Road diets are a useful tool in the traffic engineer’s arsenal of making streets and roads a more integral part of the community. As a part of broader plans, they can open up traditional roads to greater use by pedestrians and, especially, bicyclists. In general, safety benefits can be expected but vary greatly from site to site. When corrected for citywide trends, the results here indicate that only a relatively modest CMF (0.91) is appropriate. This indicates that crash/safety benefits are likely to be considerably less than what is suggested by a simple comparison of before and after crash statistics. Similarly, the results reported here also suggest that an operational analysis (e.g., using Synchro) should always be performed early on in the consideration of a road-diet proposal. Moreover, the use of the ADT for consideration of a road is not appropriate. Realistic peak-hour analyses (based on actual counts) are much more useful. For the sites evaluated here, the peak-hour threshold volume is estimated as 1,000 vph, although this could vary with different volumes of cross traffic at intersections. ST
