Advancing healthy cities through safer cycling: An examination
of shared lane markings
Nicholas N. Ferenchak
, Wesley E. Marshall
University of New Mexico, Department of Civil, Construction & Environmental Engineering, MSC 01 1070, Albuquerque, NM 87131, United States
University of Colorado Denver, Department of Civil Engineering, United States
Received 20 April 2018
Received in revised form 27 December 2018
Accepted 31 December 2018
Available online 9 January 2019
Shared lane markings
To advance healthy transportation via increased bicycling, cities combat one of the primary
barriers to such cycling – trafﬁc safety concerns – through the provision of various bicycle
treatments. Shared lane markings (more commonly known as ‘‘sharrows”) are an increas-
ingly common treatment utilized to improve bicyclist safety. While past research conﬁrms
that sharrows may effectively inﬂuence spacing and other operational measures, the
impact on actual safety outcomes remains unsubstantiated due to a lack of dooring-
related bicycle crash data. Fortunately, the city of Chicago instituted a program to collect
dooring crash data in 2010. Thus, the purpose of this research is to longitudinally examine
the association between sharrows and bicyclist injuries by combining traditional crash
data with dooring-related crash data.
To perform this examination, we divide Census block groups in Chicago into three cate-
gories based on what bicycle treatment was installed between the years 2011 and 2014: (i)
those block groups with no bicycle facilities installed; (ii) those with only sharrows
installed; or (iii) those with only bicycle lanes (standard, buffered, and/or protected)
installed. Negative binomial regressions and Kruskal–Wallis tests suggest that block groups
with only sharrows installed experienced the largest increase in bicyclist injury rates, with
exposure being accounted for through levels of bicycle commuter activity. This relationship
held true for overall crashes as well as for dooring-related crashes. These ﬁndings raise
concerns regarding the safety effectiveness of sharrows as used by the City of Chicago dur-
ing the study period and should be a call for more research on the subject in a variety of
different contexts using various exposure metrics.
Ó2019 Tongji University and Tongji University Press. Publishing Services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
Bicycling has been shown to have an overall positive impact on the health of those who ride (de Hartog et al., 2010; Rojas-
Rueda et al., 2011; Deenihan and Caulﬁeld, 2014). However, one of the primary barriers to bicycling is trafﬁc safety concerns
(Fowler et al., 2017). By improving trafﬁc safety outcomes for bicyclists, we can expect not only direct health beneﬁts in
terms of reduced injuries and fatalities, but indirect health beneﬁts accrued through greater participation and increased
2046-0430/Ó2019 Tongji University and Tongji University Press. Publishing Services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer review under responsibility of Tongji University and Tongji University Press.
E-mail address: email@example.com (N.N. Ferenchak).
International Journal of Transportation Science and Technology 8 (2019) 136–145
Contents lists available at ScienceDirect
International Journal of Transportation
Science and Technology
journal homepage: www.elsevier.com/locate/ijtst
physical activity. Research has also shown that cities with elevated levels of bicycling have better safety outcomes for all
road users and lower air pollution, further amplifying the health beneﬁts of bicycling (Marshall and Garrick, 2011;
Johansson et al., 2017). In terms of advancing healthy cities, enabling more bicycling through improved safety is a worthy
One common method of improving trafﬁc safety for bicyclists is through the implementation of bicycle treatments and
facilities. An extensive toolbox of such treatments exists, ranging from signage and wayﬁnding to dedicated and protected
facilities exclusive to bicyclists. One of the most widely used treatments is the shared lane marking (Fig. 1).
Shared lane markings – more commonly known as sharrows – trace their origins to Denver, Colorado in the early 1990s
(Alta Planning + Design, 2004; Pein et al., 1999). The markings were initially purposed to improve bicyclist safety by raising
driver awareness of bicyclists and reducing wrong-way riding (City and County of Denver, 1993). They have since evolved to
serve several different functions such as reducing sidewalk riding and avoiding collisions with the doors of parked cars (i.e.
dooring crashes) (U.S. Department of Transportation, 2009; National Association of City Transportation Ofﬁcials (NACTO),
2011). In 2009, sharrows were added to the Federal Highway Administration’s Manual on Uniform Trafﬁc Control Devices
(MUTCD), solidifying their place as an accepted bicycle treatment (U.S. Department of Transportation, 2009). The markings
have become a popular substitute for more expensive and expansive alternatives such as bicycle lanes and cycle tracks.
Today, sharrows comprise the majority of nearly every major U.S. city’s bicycle network and have become a staple in the
toolboxes of transportation planners and engineers. However, little past research has adequately examined whether these
markings help make bicyclists safer.
One of the primary reasons for this absence of safety research is the lack of data on dooring crashes. Since one of the sta-
ted justiﬁcations for sharrow installation is to help move bicyclists out of the door zone, a safety analysis without dooring
crashes would be insufﬁcient. Nearly all cities neglect dooring crashes because the crashes do not involve a moving motor
vehicle, therefore failing to meet the standard of what constitutes a motor vehicle crash. One of the only exceptions is the
City of Chicago. In 2010, the City of Chicago initiated a program to collect dooring crash data. The release of this dooring
crash data marks the ﬁrst time that the impact of sharrows on bicyclist safety can begin to be properly studied.
With their popularity rising in cities across the country, this paper aims to longitudinally examine safety outcomes – in
terms of bicyclist injuries – of Chicago block groups that had sharrows installed against block groups that installed bicycle
lanes (standard, buffered, and/or protected) as well as those that added no new bicycle facilities. More speciﬁcally, we inves-
tigate changes in bicyclist injury counts within these treatment typology groups by utilizing negative binomial regressions.
We then analyze changes in injury rates between the different typologies through a Kruskal–Wallis test.
Fig. 1. Example of a sharrow.
N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145 137
Over the past two decades, interest in bicycling has continued to increase across the United States and the health beneﬁts
of such activity have become better understood. Coinciding with this increased interest in bicycling, sharrows have become a
standard bicycle treatment. However, a void exists in the research as to how these treatments link to actual safety outcomes.
This paper will utilize spatial and statistical analyses to examine the relationship between sharrows and bicyclist injuries.
2. Literature review
Although their overall goal is to improve bicyclist safety, the exact operational function of sharrows is multifaceted and
seems to have evolved over time. Many early studies that examined sharrows identiﬁed the altering of bicycle and vehicle
spacing as an objective, including the spacing of bicyclists to avoid dooring crashes (Alta Planning + Design, 2004; U.S.
Department of Transportation, 2010; Pucher et al., 2010). Similarly, four of the MUTCD’s ﬁve objectives for sharrows deal
with lateral spacing, the ﬁrst of which is to assist bicyclists with lateral positioning to help bicyclists avoid impacting the
open door of a parked vehicle (U.S. Department of Transportation, 2009). While results are mixed, past studies suggest that
the effects of sharrows on spacing tend to be theoretically positive (Hunter et al., 2011, 2012; Brady et al., 2010). In other
words, the mean distance between bicycles and parked cars, between bicycles and the curb, and between bicycles and mov-
ing vehicles can increase up to 10.5 inches with the installation of sharrows, which gives bicyclists more space to operate,
and in theory, safer operating conditions (Hunter et al., 2011, 2012; Brady et al., 2010; Sando, 2014). However, other studies
suggest no signiﬁcant changes in lateral spacing at certain sites (Pein et al., 1999).
Today, this objective of altering spacing is less often the primary aim when installing a sharrow. Guidelines such as the
MUTCD and NACTO’s Urban Bikeway Design Guide recommend using sharrows to accomplish a variety of other operational
objectives, such as alerting road users of bicyclists’ presence, encouraging safe passing behaviors, reducing wrong-way rid-
ing, indicating the proper riding path over hazards such as railroad tracks, functioning as wayﬁnding devices, and reducing
sidewalk riding (U.S. Department of Transportation, 2009; NACTO, 2011).
The academic community has focused on the impact that sharrows have on these other operational measures as well
(Pein et al., 1999; Furth et al., 2011; Hunter et al., 2011, 2012). Overall, past studies have reported inconsistent ﬁndings
regarding the installation of sharrows in terms of weaving through vehicle queues, sidewalk riding, and wrong-way riding
(Hunter et al., 2012; Alta Planning + Design, 2004; U.S. Department of Transportation, 2011; Brady et al., 2010). While shar-
rows have been used to accomplish a wide array of operational objectives with varying levels of success, the impact on the
overall goal of improved bicyclist safety has been largely neglected (Brady et al., 2010), despite the fact that existing research
implies a direct beneﬁt to safety without examining crashes, injuries, or fatalities (Alta Planning + Design, 2004; Hunter et al.,
2011, 2012; Brady et al., 2010).
The lone piece of research that we found examining the impact that sharrows have on safety outcomes suggests that shar-
rows may be less safe than other bicycle treatments or even less safe than having no bicycle treatments present at all (Harris
et al., 2013). Although not reaching statistical signiﬁcance, results of Harris et al. (2013) suggest that sharrows increase bicy-
clists’ risk of injury at non-intersection locations while cycle tracks and bicycle lanes signiﬁcantly decrease bicyclists’ risk of
injury compared to similar roadways that lacked bicycle infrastructure. This case-crossover study, which took place in Van-
couver and Toronto, Canada, examined adult bicyclists who were injured and treated at a hospital. The researchers identiﬁed
the injury location (the case site), selected two other locations along each bicyclist’s route (control sites), recorded roadway
characteristics of the sites, and then compared the likelihood of being injured based on the different roadway characteristics.
In this way, exposure was accounted for because all sites were similarly located along the cyclists’ routes. Crashes that did
not result in a hospital stay were not included in order to better focus on safety-related health outcomes. Dooring crashes
were also not explicitly included or separately analyzed. This lack of dooring crashes presents a gap in the current literature.
Because ﬁndings from past research suggest that, by increasing spacing, sharrows should help improve safety outcomes –
especially in terms of dooring crashes – not including dooring crashes in a safety study means that one of the main hypoth-
esized beneﬁts is being excluded.
In order to explore the relationship between sharrows and bicyclist safety outcomes, we examined Chicago block groups
between 2011 and 2014 using bicyclist injury, exposure, and treatment data. We then examined the changes in bicyclist
injuries both within the individual treatment typologies and between the treatment typologies.
The City of Chicago – through the Illinois Department of Transportation’s Division of Trafﬁc Safety – provided us with
bicyclist injury data. This data included bicyclist injuries from both dooring and non-dooring crashes within the City of Chi-
cago from 2011 through 2014. Bicyclist injuries were based on police reports submitted by law enforcement agencies. The
reports provided crash locations for all injuries occurring on public property that were reported to and recorded by police.
Severity of the bicyclists’ injuries was not provided and was therefore not utilized in this study. We used injuries instead of
fatalities because of the relative scarcity of bicyclist fatalities within the City of Chicago. Originally in spreadsheet format, we
geocoded the injury data using ArcGIS and created a layer with each injury crash represented as one spatial point.
Beyond the typical absence of dooring crash data, another reason for the lack of sharrow safety research is inadequate
bicycle exposure data. An ideal exposure metric would include bicyclist counts for every segment and intersection within
138 N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145
the city and across the desired time frame. Unfortunately, no cities collect such extensive bicycling data, especially not on an
adequately wide longitudinal time frame.
An alternative approach to accounting for bicyclist exposure is through bicycle commuter data from the American Com-
munity Survey (ACS). The underlying assumption that bicycle commuters are an indicator of total bicycling exposure has
been shown by past research to be a reasonable assumption (Barnes and Krizek, 2005; Turner et al., 1997) and applicable
for bicycle safety studies (Aultman-Hall and Kaltenecker, 1999; Marshall and Garrick, 2011; Chen et al., 2012). While not
meeting the characteristics of an ideal exposure metric, ACS data allowed us to conduct this study with bicycle commuter
counts on the block group level. These journey to work data (5-year compilations) and block group boundary layers were
made available through the National Historical Geographic Information System (NHGIS) for each of the study years
(Manson et al., 2017).
We then acquired bicycle treatment data from the City of Chicago Data Portal in ArcGIS format. Because this data did not
include dates of installation, we dated the bicycle treatments by utilizing historic satellite imagery from Google Earth, his-
toric Google Street View, Chicago bicycle maps, and communications with city planners. Installation years were identiﬁed for
each roadway segment. Because we examined treatments implemented between 2011 and 2014, the available resources
were able to provide the necessary precision.
There were considerable increases in the mileage of both bicycle lanes and sharrows in Chicago between 2011 and 2014
(Table 1). These bicycle lanes and sharrows were distributed throughout the city (Fig. 2). There were 1948 block groups
within the city that had no bicycle treatments installed in 2012 or 2013, 42 block groups that had only sharrows installed,
and 149 block groups that had only bicycle lanes installed. There were also 19 block groups that had both sharrows and bicy-
cle lanes installed and 20 block groups that had bicycle lane upgrades (i.e. from standard to protected). Block groups that had
both sharrows and bicycle lanes installed were not included in the analysis, as combining the two types of treatments may
have resulted in a unique impact on safety. While these block groups could have been considered as their own category, their
rarity (only nineteen out of 2178 block groups) precluded any statistical signiﬁcance. Bicycle lane upgrade block groups were
not included in the analysis as their inclusion could confound results.
After obtaining all necessary data, we operationalized bicycle treatments on the block group level by designating each
2010 block group as one of three types: (i) block groups that had no bicycle treatments installed in 2012 or 2013; (ii) block
groups that had only sharrows installed in 2012 or 2013; or (iii) block groups that had only bicycle lanes (standard, buffered,
and/or protected) installed in 2012 or 2013. We designated our ‘before’ period as 2011–2012 and our ‘after’ period as 2013–
Using data on the block group level employs the assumption that the impacts of bicycle treatment installation are expe-
rienced throughout the entire block group. While this macro-scale approach limited the detailed exploration of the impacts
of street design and other roadway characteristics, the methodology did allow us to accomplish a city-wide analysis on the
block group level – a geographic level that has been shown to be effective for detecting differences in safety outcomes
(Marshall and Garrick, 2010; Ferenchak and Marshall, 2017) – that included dooring data and bicyclist exposure.
We derived the number of bicyclist injuries in the before and after periods by spatially joining the injury point layer to the
block group layer in ArcGIS. Because block groups are often delineated by roads, we avoided edge issues by utilizing a 50-foot
buffer around each block group. In other words, if an injury occurred on the border between two block groups, we had that
injury count for both block groups. We then joined the number of bicycle commuters in both the before and after periods for
each block group from the ACS spreadsheets. In this way, every block group included a typology (bicycle lanes, sharrows, or
Descriptive statistics for study block groups.
NMean SD Min Max
2011–2012 3060 1.4 2.6 0 41
2013–2014 3174 1.5 2.6 0 40
2011–2012 13,985 6.5 15.9 0 156
2013–2014 16,903 7.9 17.0 0 167
Mileage (2011) Mileage (2014) Block groups
Sharrows 33.5 41.4 42
Bicycle lanes 122.1 157.0 149
Standard 122.1 131.6 n/a
Buffered 0 18.2 n/a
Protected 0 7.2 n/a
None n/a n/a 1948
N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145 139
none), the number of bicyclist injuries (both dooring and non-dooring) that occurred in the before period (2011–2012) and
the after period (2013–2014), and the number of bicycle commuters in the before period and the after period.
Using the above data, we utilized negative binomial regressions to analyze the signiﬁcance of the change in bicyclist
injury counts over the study period within each individual treatment typology. Negative binomial regressions have been
shown to be an appropriate approach when examining differences in macro-level and block group level bicyclist safety rel-
ative to changes in bicycle treatments (Wei and Lovegrove, 2012; Dumbaugh and Li, 2010). The negative binomial regression
accounts for the overdispersion that we had in our data – and which is typically seen in trafﬁc crash and injury data – and is
the most appropriate and accepted practice for road safety researchers. We utilized the number of commuters and facility
typology as independent variables and injury counts as the dependent variable in our models.
Fig. 2. Blocks groups with different bicycle treatments installed in 2012 and 2013.
140 N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145
We then sought to compare the safety impacts of the different treatments in terms of injury rates. Because the analysis
called for the comparison of the changes between pre- and post-installation scenarios for three different groups, a Kruskall-
Wallis test was appropriate. This is a common approach for testing pre- and post-scenarios across multiple groups, especially
in medical literature (Brennan et al., 1997; Ratcliffe et al., 2002; Sßentürk et al., 2002). The Kruskal–Wallis test is a non-
parametric one-way analysis of variance used to compare data when the data is not normally distributed. ANOVA was
not used in this case because, although the differences among group rate means were being explored, the unbalanced sample
sizes proved problematic for this type of statistical analysis, and the data did not ﬁt the normal distribution (Shaw and
In order to complete the Kruskal–Wallis tests, we calculated bicyclist injury rates in the form of bicyclist injuries per 100
bicycle commuters. Because some block groups lacked bicycle commuters, we assumed one rider for null values in order to
avoid artiﬁcial rate inﬂation and undeﬁned values. We then weighted these rates based on the number of bicyclists in each
block group. For instance, the injury rate of a block group with 100 bicycle commuters was given more weight in the overall
average than a block group that had only 1 bicycle commuter. We used the margin of errors provided by the ACS in order to
create conﬁdence intervals around the rates. Using the Kruskal–Wallis test, we ﬁrst compared bicyclist injury rates within
individual treatment typologies and then compared the changes between the different treatment typologies.
We ﬁrst analyzed the number of bicycle commuters in relation to the type of treatment that was installed over the study
period. Block groups with only sharrows installed had the largest increase in bicycle commuters (Table 2). Block groups that
had no treatments installed had the smallest absolute increase in bicycle commuters. Both changes were signiﬁcant at 95%
conﬁdence according to paired t-tests. Block groups that had bicycle lanes installed had the smallest percentage increase,
although the change was not statistically signiﬁcant.
5.1. Injuries within block group types
Negative binomial regressions allowed us to explore changes in bicyclist injury counts that occurred within the block
group types for three categories of injuries: total injuries, dooring injuries, and non-dooring injuries (Table 3). Block groups
that had no bicycle treatments installed and block groups that had sharrows installed saw decreases in total injuries, dooring
injuries, and non-dooring injuries. All of these changes were found to be statistically signiﬁcant. For block groups that had
bicycle lanes installed, dooring injuries experienced a statistically signiﬁcant decrease. However, overall injuries experienced
a statistically signiﬁcant increase. The change in non-dooring injuries in bicycle lane block groups was not statistically
5.2. Injury rates within block group types
Ridership numbers for the different typologies enabled us to normalize bicyclist injuries and weight the injury rates to
account for changes in ridership brought about by these treatments (Table 4). Kruskal–Wallis tests examined changes in
injury rates within block group typologies between pre- and post-conditions. Because the bicycle lane block groups that
had the smallest increases (or actually saw decreases) in injury rates had the most bicycle commuters (and were therefore
weighted the heaviest), the overall injury rates in bicycle lane block groups saw the smallest percentage increase and the
second smallest absolute increase (although these changes were not statistically signiﬁcant). Similarly, block groups that
had no bicycle treatments installed had their highest injury rate increases in block groups with relatively few bicyclists. This
resulted in the smallest overall increase and second smallest percentage increase in overall injury rates (not statistically sig-
niﬁcant). Block groups that had sharrows installed had their largest injury rate increases occur in block groups that had the
most riders. For this reason, the absolute and percentage increases in bicyclist injury rates were highest in sharrow block
groups (statistically signiﬁcant at 90% conﬁdence). Dooring injury rates were reduced in block groups that had bicycle lanes
installed (statistically signiﬁcant at 90% conﬁdence) and in block groups that had no bicycle treatments installed (not statis-
tically signiﬁcant), but dooring injury rates increased in block groups with sharrows (not statistically signiﬁcant).
Change in bicycle commuters for block groups with different types of bicycle treatments installed and corresponding paired t-test results.
Bicycle commuters per
nBefore After Change % change p-value
None 1948 6.07 7.34 1.27 20.9% 0.000
Sharrow 42 10.31 15.50 5.19 50.3% 0.032
Bicycle lane 149 11.56 13.16 1.60 13.8% 0.224
N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145 141
5.3. Injury rates between block group types
The Kruskal–Wallis test also allowed us to examine the difference in injury rate changes between typologies. Did injury
rates increase more for bicycle lane, sharrow, or no-treatment block groups? Block groups that had bicycle lanes installed
and those with no treatments installed saw a statistically smaller increase in bicyclist injury rates than sharrow block groups
(Table 5). This relationship was signiﬁcant at 95% conﬁdence. Block groups that had bicycle lanes installed and block groups
that had no bicycle treatments installed did not have signiﬁcantly different changes in bicyclist injury rates.
Results suggest that in Chicago for the time period studied, block groups that had sharrows installed experienced poorer
safety outcomes than those experienced by block groups that had bicycle lanes installed or that did not install any bicycle
treatments. Block groups that had bicycle lanes installed saw the largest increase in bicyclist injuries between the before and
the after periods while block groups that had sharrows installed saw a decrease in overall bicyclist injuries. However, after
normalizing based on the increases in bicycle commuters, Chicago block groups that had bicycle lanes installed and those
that had no bicycle treatments installed experienced similarly small increases in injury rates. This is in part because bicycle
lane block groups with the largest increases in injuries had the fewest riders. Block groups that had sharrows installed expe-
rienced large increases in both the percentage change in injury rate and the absolute change in injury rate. This is because
sharrow block groups that had decreases in injury rate had few bicycle commuters, while sharrow block groups that had
increases in injury rate had many bicycle commuters. The increase in the bicyclist injury rate for Chicago block groups that
had sharrows installed was statistically greater – with 95% conﬁdence – than the increases experienced by bicycle lane block
groups or those that had no treatments installed.
Improving bicycling safety barriers can enable more people to get on their bikes and thereby unlock considerable health
beneﬁts for those bicyclists (Fowler et al., 2017; de Hartog et al., 2010; Rojas-Rueda et al., 2011; Deenihan and Caulﬁeld,
Negative binomial regression results for bicyclist injuries from before to after installation.
nB Std. error Sig.
None 1948 0.275 0.0600 0.000
Dooring 8.663 0.3550 0.000
Non-Dooring 0.204 0.0638 0.001
Sharrow 42 3.946 1.0188 0.000
Dooring 13.074 2.0787 0.000
Non-Dooring 6.143 2.4497 0.012
Bicycle Lane 149 5.416 1.7730 0.002
Dooring 8.522 1.0761 0.000
Non-Dooring 0.057 0.1767 0.747
Weighted safety rates for block groups with different types of bicycle treatments installed.
Weighted injuries per year
per 100 bicycle commuters
nBefore After Change Change % p-value
None 1948 8.70 9.75 1.05 12.1% 0.260
Dooring 1.82 1.40 0.41 22.5% 0.192
Non-Dooring 6.68 8.35 1.67 25.0% 0.041
Sharrow 42 8.97 18.28 9.31 103.8% 0.085
Dooring 1.34 4.23 2.89 215.7% 0.124
Non-Dooring 7.12 14.05 6.94 97.5% 0.121
Bicycle Lane 149 16.63 18.32 1.69 10.2% 0.986
Dooring 2.00 0.78 1.22 61.0% 0.053
Non-Dooring 14.53 17.54 3.02 20.8% 0.339
Signiﬁcance of Kruskal–Wallis test examining the longitudinal change in bicycle safety at 95% conﬁdence.
Injuries per year per 100 bicycle commuters
Smaller increase in injury rate
Bicycle lane None
Greater increase in injury rate Bicycle lane n/a 0.779
Sharrow 0.050 0.004
142 N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145
2014). While bicycle treatments are a popular method of improving safety, not all bicycle treatments are equivalently effec-
tive (Harris et al., 2013). This work begins to ﬁll an important gap in bicycle treatment research by investigating the safety
impact of sharrows. As sharrows have become an accepted treatment in cities across the United States despite a lack of safety
evaluation, this conversation has become a critical one. Our ﬁndings suggest that Chicago block groups that had sharrows
installed experienced less than desirable safety outcomes for the study period. This paper ﬁnds that there is reason to further
question the impact of sharrows on bicyclist safety outcomes through future research in other contexts.
While the mechanisms behind these ﬁndings are not yet understood and cannot be determined from this paper’s results,
one possible explanation is that the sharrows – as installed in Chicago during the study period – provide a false sense of secu-
rity to bicyclists (Hunter et al., 2000). When a bicycle lane or other separated facility is provided, the bicyclist is granted ded-
icated space. This dedicated space lowers the risk of collision with a motor vehicle (Khan and Langlois, 2011). Alternatively, if
bicycle treatments are not provided on a roadway, it is understood that the bicyclist will need to share the travel lane with
vehicles (Landis et al., 1997). The bicyclist should therefore ride in a manner appropriate to the level of risk created by the
volume, speed, and other characteristics of the roadway. When a sharrow is provided, however, bicyclists may believe that
they are at lower risk because a treatment is being provided and change their behavior accordingly. Furthermore, new bicy-
clists – possibly less experienced ones – may be attracted to the facility. While the operations (e.g. lateral spacing) of vehicles
and bicycles may be altered by the presence of the sharrow, those operational changes may not necessarily lower the prob-
ability of a collision with a motor vehicle. Might there be situations in which moving a bicyclist further in or out of a lane
does not make them objectively safer? While subjective safety may be improved, objective safety may remain static, thus
resulting in poor safety outcomes. Researchers have shown that adding crosswalks without treating underlying safety issues
may give pedestrians a false sense of security and result in increased pedestrian crash rates (Herms, 1972; FHWA, 2005).
Might sharrows be inducing the same phenomenon for bicyclists? Obtaining clarity on this issue will require further
The methodologies employed throughout this paper use a number of assumptions, the validity of which should be
explored in future research. First, this research employed the assumption that the impact of bicycle treatment installation
Fig. 3. Examples of alternative forms of sharrows (clockwise from top-left): Portland’s Sharrow Flower, a diagonal wayﬁnding sharrow, Long Beach’s Green
Sharrow, and an intersection sharrow. Photo Credit: Michael Anderson, NACTO, San Francisco Bicycle Coalition, and Eric Fischer
N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145 143
will be experienced throughout the block group within which the treatment is installed. This assumption was used primarily
so that we could perform analysis on the block group level, which allowed for the bicycle commuter metric to be utilized for
exposure. While it would be ideal to longitudinally examine speciﬁc corridors that had sharrows installed, this would have
required ridership counts speciﬁc to those corridors, which was not feasible based on the size of the study and the lack of
preexistent bike count data. A further limitation of the bicycle commuter data that we used to represent exposure is that the
data only considers the origin of the commute trip, not the destination or the path that the commuter uses. It may be that a
commuter lives on the edge of their block group and does not ride through the block group that they live in. That being said,
bicycle commuter counts have been shown to be a reasonable indicator of total bicycling exposure (Barnes and Krizek, 2005;
Turner et al., 1997) and to be applicable for bicycle safety studies (Aultman-Hall and Kaltenecker, 1999; Marshall and
Garrick, 2011; Chen et al., 2012).
In future corridor-level work, the amount of bicycle treatments could be accounted for in order to directly determine the
strength of the relationship between sharrows and bicyclist safety, as opposed to the inter-typology comparative approach of
this work. Accounting for injury severity in future work may further inform the relationship that sharrows have with bicy-
clist safety. Furthermore, future studies might beneﬁt from the incorporation of demographics or socio-economic data into
their models. Demographic and socio-economic factors have been shown to be correlated with bicycle ridership and trafﬁc
safety outcomes. Inclusion of these factors may allow for a more robust model and a better understanding of the complexity
of these issues.
The implications of this work help contribute toward the goal of bettering trafﬁc safety outcomes within our cities. The
objective of this research was to identify bicycle treatments that are effective at improving bicyclist safety. The results of the
analysis indicate that we may need to further examine exactly where and when sharrows should be used. As part of this, it
may be useful to remember that sharrows are signage and not actual bike infrastructure. Under current standards and guid-
ance, sharrows can be used for a wide variety of reasons and in a wide variety of roadway conditions: from wayﬁnding
devices to lateral spacing guides and from local roads to arterials. Researchers have not explored all of the different combi-
nations in which sharrows can and are being used. This paper does not claim to inform any of these speciﬁc scenarios but
broadly identiﬁes one case study in which safety improvements seem to be lacking. Fig. 3 shows some examples of the wide
variety of shapes, sizes, and forms in which sharrows have appeared. This work strictly evaluated the effectiveness of the
MUTCD version of the sharrow that was used in Chicago, but future work that explores different sharrow shapes, sizes,
forms, functions, and roadway scenarios would be warranted.
With sharrows becoming a familiar sight on our roadways and the health beneﬁts of bicycling becoming clearer, it is vital
to fully understand the impact that sharrows have on bicyclist safety. While past research has identiﬁed the relationship
between sharrows and operational metrics such as lateral spacing, it is only a theoretical means to an end. The effectiveness
of sharrows in terms of the true goal, which is reducing bicyclist injuries and fatalities, remains unclear in the current body of
research. This work begins to question their effectiveness and should act as a call for more research on the subject. It is
imperative that appropriate treatments are in place to ensure the safety of all users on our roadways to thereby enable
the substantial health beneﬁts that stand to be gained. It remains to be seen what role sharrows play in this pursuit.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijtst.2018.12.003.
Alta Planning + Design, 2004. San Francisco’s Shared Lane Pavement Markings: Improving Bicycle Safety. San Francisco Department of Parking & Trafﬁc, San
Aultman-Hall, L., Kaltenecker, M.G., 1999. Toronto bicycle commuter safety rates. Accid. Anal. Prev. 31, 675–686.
Barnes, G., Krizek, K., 2005. Estimating bicycle demand. Transp. Res. Rec. 1939, 45–51.
Brady, J., Loskorn, J., Mills, A., Duthie, J., Machemehl, R., 2010. Effects of Shared Lane Markings on Bicyclist and Motorist Behavior along Multi-lane Facilities.
The Center for Transportation Research, the University of Texas at Austin, Austin, Texas.
Brennan, T.J., Umali, E.F., Zahn, P.K., 1997. Comparison of pre- versus post-incision administration of intrathecal bupivacaine and intrathecal morphine in a
rat model of postoperative pain. J. Am. Soc. Anesthesiologists 87 (6), 1517–1528.
Chen, L., Chen, C., Srinivasan, R., McKnight, C.E., Ewing, R., Roe, M., 2012. Evaluating the safety effects of bicycle lanes in New York city. Am. J. Public Health
102 (6), 1120–1127.
City and County of Denver, 1993. 1993 Denver Bicycle Master Plan. Department of Public Works, Denver, Colorado.
De Hartog, J.J., Boogaard, H., Nijland, H., Hoek, G., 2010. Do the health beneﬁts of cycling outweigh the risks? Environ. Health Perspect. 118 (8), 1109–1116.
Deenihan, G., Caulﬁeld, B., 2014. Estimating the health economic beneﬁts of cycling. J. Transp. Health 1 (2), 141–149.
Dumbaugh, E., Li, W., 2010. Designing for the safety of pedestrians, cyclists, and motorists in urban environments. J. Am. Plan. Assoc. 77 (1), 69–88.
Federal Highway Administration, 2005. Safety Effects of Marked Versus Unmarked Crosswalks at Uncontrolled Locations. HRT-04-100. U.S. Departmentof
Transportation, McLean, VA.
Ferenchak, N.N., Marshall, W.E., 2017. Redeﬁning the child pedestrian safety paradigm: identifying high fatality concentrations in urban areas. Injury
Prevent. 23 (6), 364–369.
Fowler, S.L., Berrigan, D., Pollack, K.M., 2017. Perceived barriers to bicycling in an urban U.S environment. J. Trans. Health 6, 474–480.
Furth, P.G., Dulaski, D.M., Bergenthal, D., Brown, S., 2011. More than sharrows: lane-within-a-lane bicycle priority treatments in three U.S. cities.
Transportation Research Board 2011 Annual Meeting. Washington, D.C.
Harris, M.A., Reynolds, C.C.O., Winters, M., Cripton, P.A., Shen, H., Chipman, M.L., Cusimano, M.D., Babul, S., Brubacher, J.R., Friedman, S.M., Hunte, G., Monro,
M., Vernich, L., Teschke, K., 2013. Comparing the effects of infrastructure on bicycling injury at intersections and non-intersections using a case-
crossover design. Injury Prevent. 19, 303–310.
144 N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145
Herms, B.F., 1972. Pedestrian crosswalk study: accidents in painted and unpainted crosswalks. Highway Res. Rec. 406, 1–13.
Hunter, W.W., Harkey, D.L., Stewart, J.R., Birk, M.L., 2000. Evaluation of blue bike-lane treatment in Portland, Oregon. Transp. Res. Rec. 1705, 107–115.
Hunter, W.W., Srinivasan, R., Martell, C.A., 2012. Evaluation of Shared Lane Markings in Miami Beach, Florida. University of North Carolina, Highway Safety
Hunter, W.W., Srinivasan, R., Thomas, L., Martell, C.A., Seiderman, C.B., 2011. Evaluation of shared lane markings in Cambridge, Massachusetts. Transp. Res.
Rec. 2247, 72–80.
Johansson, C., Lövenheim, B., Schantz, P., Wahlgren, L., Almström, P., Markstedt, A., Strömgren, M., Forsberg, B., Sommar, J.N., 2017. Impacts on air pollution
and health by changing commuting from car to bicycle. Sci. Total Environ. 584–585, 55–63.
Khan, A.M., Langlois, R.G., 2011. Factoring cycling in transportation infrastructure: design considerations based on risk exposure. ITE J. 81 (8), 49–53.
Landis, B.W., Vattikuti, V.R., Brannick, M.T., 1997. Real-time human perceptions: toward a bicycle level of service. Transp. Res. Rec. 1578, 119–126.
Manson, S., Schroeder, J., Van Riper, D., Ruggles, S., 2017. IPUMS National Historical Geographic Information System: Version 12.0 [Database]. University of
Marshall, W.E., Garrick, N.W., 2010. Street network types and road safety: a study of 24 California cities. Urban Des. Int. 15 (3), 133–147.
Marshall, W.E., Garrick, N.W., 2011. Evidence on why bike-friendly cities are safer for all road users. Environ. Pract. 13 (1), 16–27.
National Association of City Transportation Ofﬁcials. (2011). Urban bikeway design guide. April 2011 Edition.
Pein, W.E., Hunter, W.W., Stewart, J.R., 1999. Evaluation of the Shared-use Arrow. Florida Department of Transportation, Pedestrian/Bicycle Safety Section.
Pucher, J., Dill, J., Handy, S., 2010. Infrastructure, programs, and policies to increase bicycling: an international review. Prev. Med. 50, S106–S125.
Ratcliffe, J., Longworth, L., Young, T., Bryan, S., Burroughs, A., Buxton, M., 2002. Assessing health-related quality of life pre- and post-liver transplantation: a
prospective multicenter study. Liver Transpl. 8 (3), 263–270.
Rojas-Rueda, D., de Nazelle, A., Tainio, M., Nieuwenhuijsen, M.J., 2011. The health risks and beneﬁts of cycling in urban environments compared with car
use: health impact assessment study. BMJ 343, d4521.
Sando, T., 2014. Operational Analysis of Shared Lane Markings and Green Bike Lanes on Roadways with Speeds Greater than 35 mph. The Florida
Department of Transportation Research Center, Tallahassee, Florida.
Sßentürk, M., Özcan, P.E., Talu, G.K., Kiyan, E., Çamci, E., Özyalçin, S., Dilege, S., Pembeci, K., 2002. The effects of three different analgesia techniques on long-
term postthoracotomy pain. Anesth. Analg. 94 (1), 11–15.
Shaw, R.G., Mitchell-Olds, T., 1993. ANOVA for unbalanced data: an overview. Ecology 74 (6), 1638–1645.
Turner, S., Hottenstein, A., Shunk, G., 1997. Bicycle and Pedestrian Travel Demand Forecasting: Literature review. Texas Transportation Institute, College
U.S. Department of Transportation, 2009. Manual on Uniform Trafﬁc Control Devices for Streets and Highways: 2009 Edition. Federal Highway
Administration, McLean, Virginia.
U.S. Department of Transportation, 2010. Evaluation of Shared Lane Markings. Federal Highway Administration, Turner-Fairbank Highway Research Center,
McLean, Virginia. FHWA-HRT-10-041.
U.S. Department of Transportation, 2011. Evaluation of Pedestrian and Bicycle Engineering Countermeasures: Rectangular Rapid-ﬂashing Beacons, HAWKs,
Sharrows, Crosswalk Markings, and the Development of an Evaluation Methods Report. Federal Highway Administration, Turner-Fairbank Highway
Research Center, McLean, Virginia. FHWA-HRT-11-039.
Wei, F., Lovegrove, G., 2012. An empirical tool to evaluate the safety of cyclists: community based, macro-level collision prediction models using negative
binomial regression. Accid. Anal. Prev. 61, 129–137.
N.N. Ferenchak, W.E. Marshall / International Journal of Transportation Science and Technology 8 (2019) 136–145 145