Residential Street Typology and Injury Accident Frequency

Residential Street Typology and Injury Accident Frequency
Peter Swift, P. E., Dan Painter, AICP, Matthew Goldstein1







Originally presented at the Congress for the New Urbanism, Denver, Co., June, 1997
Additional data added in the summer and fall of 2002
VMT introduced Summer 2006

1 Mr. Swift is the Director of Town Planning, Mid Atlantic Enterprises, LLC, Erbil, KRG, Iraq. Mr. Painter was the
Principal Transportation Planner for the City of Longmont, Colorado. Mr. Goldstein was an intern with Swift and
Associates.

---

Residential Street Typology and Injury Accident Frequency

ABSTRACT

Communities all across the U.S. are concerned about the safety of their residential streets.
Although this concern is nearly universal, the literature offers few precedents and little
information on the relative safety of common residential street typologies. This study offers a
method for analyzing the theory that the physical design of streets impacts safety. Through
research, systematic observation, and statistical analysis, an attempt was made to identify the
safest residential street form with respect to several physical characteristics. These findings
expose issues that need to be addressed by practitioners and policy-makers and encourage further
study of related topics.

Approximately 20,000 police accident reports from the City of Longmont, Colorado, were
reviewed and compared against five criteria for evaluating the probability that street design
contributed to the accidents. Once catalogued and entered into a database, each accident
location was mapped and described by several physical characteristics. To compare injury
accidents per mile per year against other factors, several correlations were explored. The most
significant relationship to injury accidents was found to be street width. AS street widths widen,
accidents per mile per year increases exponentially, and the safest residential street width are the
narrowest (curb face).

I. INTRODUCTION

A fundamental element in the field of planning and urban design is the assumption that narrow
streets in a grid pattern are safer than wider streets. Although this theory has been discussed at
length by practitioners and academics, little substantive research is available to support this
assumption. With the cooperation of the City of Longmont, Colorado, this study was designed
to complement the literature by evaluating residential street typology and related injury-accident
frequency.

Longmont, Colorado is located approximately 35 miles north of Denver and approximately 15
miles northeast of Boulder. It has experiencing a sustained period of growth as many
communities are along Colorado’s front range. Longmont has about 19.2 square miles of
annexed land and a current population of approximately 72,000.

The rate of urban growth and increasing traffic congestion are the two most often mentioned ills
in Longmont, although the quality of new development is rising on this list. Longmont has two
developments that are considered to be of the neo-traditional design. These developments with
narrow streets, alleys and well-developed street grids give the city staff and residents first hand
examples of a design that builds in a higher quality of life. However, the debate over acceptable
street widths continues in Longmont as it does in communities throughout the U.S.

II. DATA GATHERING AND MAPPING

---

The data used in this study were obtained by reviewing approximately 20,000 Longmont
accident reports. Each report was examined against several criteria to ascertain that the accident
occurred as a result of street typology. Eliminated from the study were accident reports that
included the following information;

1. Road conditions that were wet, icy or snow covered

2. Substance abuse - Any notation of the driver being impaired or suspicion of being
under the influence of any substance.

3. Traffic volume - accidents which occurred on any street other than a “local”. A local
street has less than 2,500 Average Daily Traffic (ADT).

The Colorado Department of Transportation has established an inventory of streets in a report
entitled “Longmont Street Inventory”. This includes an inventory of all the streets in the city
with respect to its length, primary surface width, through lane length and other data. The primary
surface width and total miles of street were of the greatest interest to this study. Many
discrepancies in street width were discovered. Field measurements were conducted to correct the
data. These field observations also included evaluation of the other variables in the study.

III. ACCIDENT LOCATION OBSERVATION

Thirteen physical characteristics of each accident location were systematically observed and
listed as follows;

Characteristic


Normative Value

1
Street Curves

Degree of curvature

2
Street width (curb face)
In 2' increments

3



Distance to nearest curb-cut <20', 21’-50' or >50'

4

Curb type


modified, 6" vertical or none

5 Tree density

Trees per 100 feet within 200
feet of the accident site




6



Traffic control

device distance from accident
<20', 21-50', 51-100' or >100'

7
ADT


Ave. annual vehicles per day
8
Sight
distance
clear
or
obstructed
view

---

9
Parking Density

Vehicles per 100 feet parked
within 200 feet of the
accident site.

10
VMT
Vehicle
miles
traveled

The ADT of some road segments were not known. In these cases, the ADTs were estimated by
identifying the probable travel shed and multiplying the number of houses within this area by 10.
Although the use of 10 trips per dwelling unit may be higher than the recognized trip generation
rate for attached units, this method allowed for other trips that may have used the street but had
neither an origin or destination within the travel shed.

IV. Statistical analysis and correlations

A multiple regression analysis was applied to the data with Kwikstat 4, a windows driven
statistical program. The number of accidents per mile per year (a/m/y) was used as the dependant
variable in the analysis. The results are presented in Table 1 for the 10 physical characteristics as
the independent variables.

The test for those independent variables in Table 1 having a lineal relationship is indicated by a
p(2) tail value <0.05. The most statistically significant variable to (a/m/y) was street width
(p[2]=0.0000) followed by ADT and VMT.

TABLE 1
KWIKSTAT 4 04-27-2002
-------------------------------------------------------------------------------
Multiple Linear Regression C:\STATS\DBF\RAW2.dbf
-------------------------------------------------------------------------------
Dependent variable is Acc/Mile/Year 10 independent variables, 304 cases.
------------------------------------------------------------------------------------------------
Variable
Coefficient
St. Error
t-value p(2 tail)
-------------------------------------------------------------------------------------------------
Intercept
-.9246872
0.2146736
-4.307411 0.0000
Deg. Curvature
-.0024117
0.0016678
-1.446023 0.1492
------------------------------------------------------------------------------------------------
Variable
Coefficient
St. Error
t-value p(2 tail)
Deg. Curvature
-.0024117
0.0016678
-1.446023 0.1492
Street Width
0.0387544
0.0044672
8.6752882 0.0000
Curb Cut Freq.
-.0062726
0.0118874
-.5276671 0.5981
Curb Type
0.0460912
0.0466812
0.9873608 0.3243
Tree Freq.
-.0148623
0.0282028
-.5269799 0.5986
Traf. Cont. Device -.0621405
0.0178682
-3.477706 0.0606
ADT

0.2429763
0.0353721
6.8691462 0.0000
Sight Dist.
-.0382670
0.0477721
-.8010330 0.4238
Parking Density -.0369261
0.0330313
-1.117913 0.2645
VMT

-.0000056 5.7D-07
-9.819408 0.0000
-------------------------------------------------------------------------------

Source Sum of Sqs df Mean Sq F p-value
-------------------------------------------------------------------------------
Regression 56.697728 10 5.6697728 33.214030 0.0000
Error 50.016316 293 0.1707042
-------------------------------------------------------------------------------
Total 106.71404 303

---

A thorough treatment of the data was conducted by applying quadratic, parabolic, power and
exponential regression analysis to the data for street widths of 14 to 50 feet vs. A/m/y with the
following results;

y=0.0014e(0.1323x)
Where;


Y = Accidents per mile per year

X = Street width (curb face) in feet

The corrected R2 value is 0.73.

The data for the development of the regression curve and Figure 1 are provided in table 2.


TABLE 2

Regressio
Street

n
Width
a/m/y
a/m/y
14
0.006
0.01
15
0.006
0.01
17
0.029
0.01
20
0.034
0.02
24
0.011
0.03
26
0.029
0.04
30
0.45
0.07
32
0.203
0.10
34
0.04
0.12
36
0.213
0.16
38
0.481
0.21
40
0.194
0.27
42
0.057
0.36
44
1.441
0.46
46
0.317
0.60
48
0.731
0.79
50
1.775
1.02


Using this regression equation, a typical 36-foot wide residential street has 0.16 (a/m/y) as
opposed to 0.03 for a 24 foot wide street. This difference is about a 487 percent increase in
accident rates. The data suggests that the wider the street, the greater the accident rate (see
Figure 1, below).

---

Figure 1


The statistical data also suggests that there is a linear relationship between a/m/y and both ADT
and VMT. A regression analysis of those data is somewhat conflicting, however. The ADT data
depicted in Figure 2, with a low R2 value of 0.17, has a best fit with a 5th order polynomial but is
irregular, at best. It appears that the accident rate increases with ADT to about 1,000 ADT, then
decreases.

Figure 2




Figure 3



The VMT data (Figure 3) also has a low R2 of 0.21 and suggests that accident rates decrease
with vehicle miles traveled. Although there seems to be a correlation of volumes and vehicle

---

miles traveled, the results are unclear with rather weak correlations. A regression exercise was
then run for accidents/year/1000/VMT to determine if there was significance to street width. The
r2 values were not significant for data sets greater than 36 feet of street width. There was
significance detected for streets 36 feet in width and less as shown in the following Figure 4;

For up to 71 data points, the R2
Median Accident Rate and Street Width
limit was 0.330. The R2 value of
0.4725 is, therefore, statistically
2.5
y = 0.0045e0.1406x
significant. This result correlates
R2 = 0.4725
with the earlier observation that
2
there is a significant increase in
injury accident rates for
T
D
residential streets up to 36 feet
1.5
Series1
in width. It would be safe to
/1000A
Expon. (Series1)
assume that this trend continues
ear
1
for wider streets.
c
c/Y

A
The authors then explored
0.5
emergency response activity
with regard to fire access and
street width. An evaluation of
0
0
10
20
30
40
structure fire incidents revealed
that there were two that
Street Width (ft)
occurred in the study period.
The first was in a fairly new
subdivision with a street width of 36 feet and the second was in the older part of town having a
street width of 24 feet and a rear alley (20’ right-of-way, 12’ paved surface). An interview with
the Fire Chief indicated that there was not a problem with access to either incident and response
times were adequate. The attack strategy for the second incident included access from the alley
and it was viewed as an advantage to have a second point of access in a narrow street
environment. Additionally, there were no problems directing the ladder truck to the alley as part
of the defense strategy.

V. CONCLUSION

Clear relationships are evident between accident frequency and street width. The findings
support the premise that narrower, so called “skinny” streets, are safer than standard width local
streets.

A larger question of public safety concerns fire apparatus and emergency vehicle access with
narrow streets. The service reports from the Fire Department of the City of Longmont were
evaluated for the study period. No fire related injuries or accidents occurred during the eight-
year period of the study. Additionally, there were no access or response time problems reported.

Lastly, additional research is encouraged to verify these results. In a very limited search of the
literature, three studies were noted. In the first report, the mean free speed of cars in suburban

---

roads increases linearly with the roadway width, particularly between 17 and 37 feet2. The
second paper by Giese, et al, suggests that spatial enclosure, sight distance and [width]
constriction techniques influence vehicle speeds3. More fully, vehicle speeds decrease with
width constriction. The third study indicates that building enclosure reduces vehicular speeds4.
This supports the conclusion of this study that narrower streets are safer.

It is notable that there were several physical elements that do not affect accident rates; parking
and tree density. These are also debated design elements in narrow street environments.

Finally, it is the conclusion of the authors that, because of the fire access needs, narrow streets
should not be used without at least a second means of access. This can be accomplished with
alleys and/or an interconnected network of streets.

END


2 Farouki, Omar, and William Nixon. 1976. The Effect of the Width of Suburban Roads on the Mean Free Speed of
Cars. Traffic Engineering and Control 17,2: 508-9
3 Giese, Joni L., Gary A. Davis and Robert D. Sykes, The relationship between residential street design
and pedestrian safety, Institute of Transportation Engineers Compendium of Technical Papers on CD-
ROM. August, 1997.

4 Smith, D. T. and Donald Appleyard, Improving the Residential Street Environment-Final Report, Federal
Highway Administration, Washington D. C., 1981, p. 127

---