FW: Pedalers in Peril: San Francisco's Most Dangerous Streets for Cyclists - The Bay Citizen
23 views
Skip to first unread message
Jim Baross
Feb 8, 2011, 11:23:30 AM2/8/11
to cabo...@googlegroups.com
Some interesting data.
My skim didn’t unearth any surprises – people seem to be still too slow to learn how to share public road space lawfully and courteously.
Will your review find anything specific that would help us improve bicycling safety and effectiveness?
Jim Baross
-----Original Message-----
From: James Lombardo SR.
Sent: Tuesday,
February 08, 2011 6:34 AM
http://www.baycitizen.org/bikes/story/san-francisco-bike-accidents/?sms_ss=email&at_xt=4d513d1a8d56451b%2C0
Sauerwald Mark
Feb 8, 2011, 11:53:12 AM2/8/11
to cabo...@googlegroups.com, jimb...@cox.net
| One interesting thing: They state that from 2009 to 2010, accidents increased faster than ridership - ridership increased by 3% while accidents increased by 8%. There were a couple of factors for 2010 which might be responsible, but one is the fact that due to a lawsuit, San Francisco added no bicycle infrastructure from 2006 through Mid 2010. In mid 2010 the ban was lifted and a lot of new bike lanes were added. One question which hasn't been answered is if the new bike lanes are increasing the accident rate. This will be easier to answer with 2011 data. Mark --- On Tue, 2/8/11, Jim Baross <jimb...@cox.net> wrote: |
|
Daniel Gutierrez
Feb 8, 2011, 2:45:26 PM2/8/11
to mark_sa...@yahoo.com, CABOforum
Mark,
Traffic deaths and injuries have always had a non-linear relationship for motoring and bicycling, this is why the whole "Safety in Numbers" (SIN) belief system (add more cyclists and everyone is safer) seems so intuitive to most people and looks plain wrong to someone that knows the traffic crash literature.
Smeed's law, an increasing power law curve fit to the data (not based on any causation) "showed" that increasing registered vehicle numbers (a proxy for VMT = vehicle miles total), correlated with a an increasing death rate, but sub linear (thus the power exponent is positive, but less than 1). This curve fit only worked in the period from the 1920 to 1950s, afterwards, the power decreased to zero (in the mid '70s) and went negative thereafter. So as VMT increases the total number of deaths is decreasing, and one sees this across all developed countries. The SIN claim is that an increasing power law, similar to Smeed's curve fit, is "good" because it's better than linear scaling. However, we know that traffic deaths never scaled linearly, so the linear (epidemiological) assumption is invalid, yet those in public health/public safety professionals do not realize they are making a specious comparison. They should be asking why bicycling deaths scale like the early high death count per VMT days of motoring, instead of the declining death rate per additional VMT mow experienced. In addition, researchers have used time series data of deaths versus time and VMT vs time to show trends. To understand the peak that occurred in the 1970s, one can see the peaked time series function as a product of two effects, one from the saturating growth in VMT and the other a "mysterious" exponential decay:
Full Report:
TRENDS AND FORECASTS IN MOTOR VEHICLE KILOMETRAGE, ROAD SAFETY AND ENVIRONMENTAL QUALITY
Matthijs J. Koornstra
SWOV Institute for Road Safety Research
P.O. Box 170, 2260 AD, Leidschendam, The Netherlands
Relevant Excerpt (quoted text is black and between rows of equal signs):
===============================
...The explanation of the rise and fall in fatalities has troubled researchers for a long time, asking what has caused the turn in development. The reason for a single peaked macroscopic development of fatalities, however, can be easily understood if one realizes that the fatality rate, that is the annual road fatalities per annual kilometrage, tend to decrease by annual reduction percentages that fluctuates around a constant percentage. This implies an underlying exponential decay function for the main trend in the fatality rate. As a consequence of saturating growth and constant risk decay, somewhere in time there is a turning point where the proportional reduction of the fatality rate becomes larger than the diminishing proportional increase of the traffic growth. Before that point in time the main trend in fatalities will be increasing and thereafter decreasing, because the product of the S-shaped trend for motor vehicle kilometres and the exponential decay for fatality rates gives by definition a single peaked trend of fatalities as shown in figure 1.
Figure 1. The basic model: saturating growth and exponential decay for adverse outcome rates.
Therefore, nothing else than a saturating traffic growth and an exponentially decreasing fatality rate are needed for an explanation of the single peaked development in fatalities that has been observed for the macroscopic development of road fatalities in many countries (Oppe, 1991a). Although the underlying macroscopic developments in vehicle kilometres are mainly described by a saturating S-shaped curve, economic developments may temporary deter and accelerate the development in traffic growth. Analyses of data from several countries have shown that the deviations from their underlying S-shaped macro-trends can be modelled by cyclic modifications around the macroscopic S-shaped curves for traffic growth. There also are related cyclic deviations from exponential risk decay to observe. These cyclic deviation patterns around the Sshaped trend for growth as well as around the exponential decay for the fatality rate (risk adaptation) may cause irregularities around that underlying single peaked development of the road fatalities, which may cause several smaller local peaks around the main u nderlying peak or two peaks around a repressed underlying main peak or only may increase the underlying main peak due to the different locations of the deviation cycles with relative high weights in the predictions of growth and risk. As shown in the sequel such local peaks are to be observed for the USA, while for The Netherlands the cyclic deviations only increase the single peaked development of fatalities. The analysis for Japan (Koornstra, 1993) yields two marked peaks that mask the single peaked main trend in Japan.
=========================
Here is the US Data:
=========================
====================================
So take the first graph above multiply it by the second to arrive at the peaked function in the third graph of fatalities versus time.
Here is the Netherlands Data:
====================================
====================================
This shows the same basic peaked non-linear function as the US and nearly every other developed country.
The Safety in Numbers believers think the left side of the peaked graph of fatalities (injuries) versus time is good (compared to linear scaling), because they are completely ignorant of actual crash time series data trends, so they don't realize that bike crashes following the SIN positive exponent are like the bad old days of motoring, instead of like the present negative exponent. This why when I hear someone talk about crashes scaling sub linearly (but with a positive instead of negative exponent) as being a good thing, I think, "BS!", because they have deluded themselves into thinking positive sub-linear scaling is good, because they mistakenly expect linear behavior. Linear scaling is fantasy-land, it is not the reality for traffic. Disease exposure scales linearly, traffic exposure does not!! And NEVER did!!
With this preface out of the way, onward to Mark's point: "One interesting thing: They state that from 2009 to 2010, accidents increased faster than ridership - ridership increased by 3% while accidents increased by 8%. There were a couple of factors for 2010 which might be responsible, but one is the fact that due to a lawsuit, San Francisco added no bicycle infrastructure from 2006 through Mid 2010. In mid 2010 the ban was lifted and a lot of new bike lanes were added. One question which hasn't been answered is if the new bike lanes are increasing the accident rate. This will be easier to answer with 2011 data."
So with the above digression in mind, an increase in crashes over the increase in mode share (assuming mode share is a valid proxy for exposure, and even this can be debated), would indicate a super-linear relationship, which is stratospheric compared to motoring. I agree that the 2011 data should prove interesting. This post is just the tip of the iceberg of confusion bicycling advocates and public health officials have regarding SIN.
Suffice it to say that adding more cyclists does NOT magically make them safer, so thinking that adding bikeways increases safety through encouragement is delusional thinking. Sadly common but nonetheless delusional.
Here are two images (publicly available FaceBook links) that show some interesting Long Beach data. You need to read the text in the images and the captions to understand that these data are a counterexample to the Safety In Numbers hypothesis:
<https://www.facebook.com/photo.php?pid=1237059&l=8f1b5afee6&id=1574017310>
<https://www.facebook.com/photo.php?pid=1237198&l=590a7e9843&id=1574017310>
This subject is interesting and nowhere near as simple as many think. I could write a novel about this subject (and Bob Shanteau and I have a couple of papers essentially ready for publication), so hopefully this is enough to get you on the right track for understanding the difficulties of bicycle crash trend analysis.
- Dan Gutierrez -
Long Beach, CA
(562) 244-4145 Cell
(310) 336-3075 Office
(800) 616-4714 Pager
Dan.Gu...@Charter.Net
Organizational Affiliations
Local:
Long Beach Cyclists, Founder and Technical Advisor
Aerospace Cycling Club, Founder and Current President
SouthBay Westside Transportation Mgmt. Assoc., Board Member
State:
CA Assoc. of Bicycling Organizations (CABO), District 7 Director
CABO Education Committee Co-Chair http://www.cabobike.org/
Caltrans District 7 Bicycle Advisory Committee, Policy Chair
National:
League of American Bicyclists (LAB), Certified Instructor, LCI #962
http://www.bikeleague.org/
Dual Chase Productions LLC, Co-Creator http://www.dualchase.com/
Dual Chase video hosting at Cyclist View http://www.cyclistview.com/
YouTube Channel: http://www.youtube.com/user/CyclistLorax
Traffic deaths and injuries have always had a non-linear relationship for motoring and bicycling, this is why the whole "Safety in Numbers" (SIN) belief system (add more cyclists and everyone is safer) seems so intuitive to most people and looks plain wrong to someone that knows the traffic crash literature.
Smeed's law, an increasing power law curve fit to the data (not based on any causation) "showed" that increasing registered vehicle numbers (a proxy for VMT = vehicle miles total), correlated with a an increasing death rate, but sub linear (thus the power exponent is positive, but less than 1). This curve fit only worked in the period from the 1920 to 1950s, afterwards, the power decreased to zero (in the mid '70s) and went negative thereafter. So as VMT increases the total number of deaths is decreasing, and one sees this across all developed countries. The SIN claim is that an increasing power law, similar to Smeed's curve fit, is "good" because it's better than linear scaling. However, we know that traffic deaths never scaled linearly, so the linear (epidemiological) assumption is invalid, yet those in public health/public safety professionals do not realize they are making a specious comparison. They should be asking why bicycling deaths scale like the early high death count per VMT days of motoring, instead of the declining death rate per additional VMT mow experienced. In addition, researchers have used time series data of deaths versus time and VMT vs time to show trends. To understand the peak that occurred in the 1970s, one can see the peaked time series function as a product of two effects, one from the saturating growth in VMT and the other a "mysterious" exponential decay:
Full Report:
TRENDS AND FORECASTS IN MOTOR VEHICLE KILOMETRAGE, ROAD SAFETY AND ENVIRONMENTAL QUALITY
Matthijs J. Koornstra
SWOV Institute for Road Safety Research
P.O. Box 170, 2260 AD, Leidschendam, The Netherlands
Relevant Excerpt (quoted text is black and between rows of equal signs):
===============================
...The explanation of the rise and fall in fatalities has troubled researchers for a long time, asking what has caused the turn in development. The reason for a single peaked macroscopic development of fatalities, however, can be easily understood if one realizes that the fatality rate, that is the annual road fatalities per annual kilometrage, tend to decrease by annual reduction percentages that fluctuates around a constant percentage. This implies an underlying exponential decay function for the main trend in the fatality rate. As a consequence of saturating growth and constant risk decay, somewhere in time there is a turning point where the proportional reduction of the fatality rate becomes larger than the diminishing proportional increase of the traffic growth. Before that point in time the main trend in fatalities will be increasing and thereafter decreasing, because the product of the S-shaped trend for motor vehicle kilometres and the exponential decay for fatality rates gives by definition a single peaked trend of fatalities as shown in figure 1.
Figure 1. The basic model: saturating growth and exponential decay for adverse outcome rates.
Therefore, nothing else than a saturating traffic growth and an exponentially decreasing fatality rate are needed for an explanation of the single peaked development in fatalities that has been observed for the macroscopic development of road fatalities in many countries (Oppe, 1991a). Although the underlying macroscopic developments in vehicle kilometres are mainly described by a saturating S-shaped curve, economic developments may temporary deter and accelerate the development in traffic growth. Analyses of data from several countries have shown that the deviations from their underlying S-shaped macro-trends can be modelled by cyclic modifications around the macroscopic S-shaped curves for traffic growth. There also are related cyclic deviations from exponential risk decay to observe. These cyclic deviation patterns around the Sshaped trend for growth as well as around the exponential decay for the fatality rate (risk adaptation) may cause irregularities around that underlying single peaked development of the road fatalities, which may cause several smaller local peaks around the main u nderlying peak or two peaks around a repressed underlying main peak or only may increase the underlying main peak due to the different locations of the deviation cycles with relative high weights in the predictions of growth and risk. As shown in the sequel such local peaks are to be observed for the USA, while for The Netherlands the cyclic deviations only increase the single peaked development of fatalities. The analysis for Japan (Koornstra, 1993) yields two marked peaks that mask the single peaked main trend in Japan.
=========================
Here is the US Data:
=========================
====================================
So take the first graph above multiply it by the second to arrive at the peaked function in the third graph of fatalities versus time.
Here is the Netherlands Data:
====================================
====================================
This shows the same basic peaked non-linear function as the US and nearly every other developed country.
The Safety in Numbers believers think the left side of the peaked graph of fatalities (injuries) versus time is good (compared to linear scaling), because they are completely ignorant of actual crash time series data trends, so they don't realize that bike crashes following the SIN positive exponent are like the bad old days of motoring, instead of like the present negative exponent. This why when I hear someone talk about crashes scaling sub linearly (but with a positive instead of negative exponent) as being a good thing, I think, "BS!", because they have deluded themselves into thinking positive sub-linear scaling is good, because they mistakenly expect linear behavior. Linear scaling is fantasy-land, it is not the reality for traffic. Disease exposure scales linearly, traffic exposure does not!! And NEVER did!!
With this preface out of the way, onward to Mark's point: "One interesting thing: They state that from 2009 to 2010, accidents increased faster than ridership - ridership increased by 3% while accidents increased by 8%. There were a couple of factors for 2010 which might be responsible, but one is the fact that due to a lawsuit, San Francisco added no bicycle infrastructure from 2006 through Mid 2010. In mid 2010 the ban was lifted and a lot of new bike lanes were added. One question which hasn't been answered is if the new bike lanes are increasing the accident rate. This will be easier to answer with 2011 data."
So with the above digression in mind, an increase in crashes over the increase in mode share (assuming mode share is a valid proxy for exposure, and even this can be debated), would indicate a super-linear relationship, which is stratospheric compared to motoring. I agree that the 2011 data should prove interesting. This post is just the tip of the iceberg of confusion bicycling advocates and public health officials have regarding SIN.
Suffice it to say that adding more cyclists does NOT magically make them safer, so thinking that adding bikeways increases safety through encouragement is delusional thinking. Sadly common but nonetheless delusional.
Here are two images (publicly available FaceBook links) that show some interesting Long Beach data. You need to read the text in the images and the captions to understand that these data are a counterexample to the Safety In Numbers hypothesis:
<https://www.facebook.com/photo.php?pid=1237059&l=8f1b5afee6&id=1574017310>
<https://www.facebook.com/photo.php?pid=1237198&l=590a7e9843&id=1574017310>
This subject is interesting and nowhere near as simple as many think. I could write a novel about this subject (and Bob Shanteau and I have a couple of papers essentially ready for publication), so hopefully this is enough to get you on the right track for understanding the difficulties of bicycle crash trend analysis.
- Dan Gutierrez -
Long Beach, CA
(562) 244-4145 Cell
(310) 336-3075 Office
(800) 616-4714 Pager
Dan.Gu...@Charter.Net
Organizational Affiliations
Local:
Long Beach Cyclists, Founder and Technical Advisor
Aerospace Cycling Club, Founder and Current President
SouthBay Westside Transportation Mgmt. Assoc., Board Member
State:
CA Assoc. of Bicycling Organizations (CABO), District 7 Director
CABO Education Committee Co-Chair http://www.cabobike.org/
Caltrans District 7 Bicycle Advisory Committee, Policy Chair
National:
League of American Bicyclists (LAB), Certified Instructor, LCI #962
http://www.bikeleague.org/
Dual Chase Productions LLC, Co-Creator http://www.dualchase.com/
Dual Chase video hosting at Cyclist View http://www.cyclistview.com/
YouTube Channel: http://www.youtube.com/user/CyclistLorax
Sauerwald Mark
Feb 8, 2011, 4:51:41 PM2/8/11
to Daniel.A....@aero.org, CABOforum
| Dan My thinking on this, was that looking only at the 2009 - 2010 data, there was a 3% increase in ridership, and an 8% increase in accidents. The numbers are large enough that these are probably statistically significant. My understanding of the SIN argument is essentially as you have presented it, but for small incremental changes in ridership the relationship should be essentially linear, even if you were a SIN evangelist (and I would claim that a 3% change is a small incremental increase). The fact that the accident rate increased 267% faster than the ridership rate tells me that there is something else going on causing a departure from the linear (or close to linear) relationship. My feeling is that there must have been something else that led to the step increase in accident
rate, so I was trying to consider what step changes were seen in San Francisco in 2010 - the two that came to mind were weather (2010 was unusually wet), and the change in legal situation which led to a lot of new bicycle infrastructure being added all at the same time. Mark |
Richard Masoner
Feb 8, 2011, 5:36:16 PM2/8/11
to mark_sa...@yahoo.com, Daniel.A....@aero.org, CABOforum
On Tue, Feb 8, 2011 at 1:51 PM, Sauerwald Mark <mark_sa...@yahoo.com> wrote:
The fact that the accident rate increased 267% faster than the ridership rate tells me that there is something else going on causing a departure from the linear (or close to linear) relationship.
554 reported accidents in 2009 vs 593 in 2010 -- not tiny numbers, but maybe small enough that even small perterbations result in big changes percentage wise?
Richard Masoner
Bob Sutterfield
Feb 11, 2011, 4:23:59 PM2/11/11
to CABOforum
LAB likes the Minneapolis SIN report:
http://www.bikeleague.org/blog/2011/02/ridership-up-crashes-down-safety-in-numbers-in-minneapolis/
http://www.bikeleague.org/blog/2011/02/ridership-up-crashes-down-safety-in-numbers-in-minneapolis/
Reply all
Reply to author
Forward
0 new messages