Friday 10 August 2012

Car vs. Bus vs. Bicycle



Just noticed this on John Allen's blog from a few years ago. John Allen has some great criticisms of it, but I had to repost the image, as I think it's a great illustration of the potential solutions to gridlock.

Of course, I say 'potential solutions', because one of the problems with the bicycles in the above image is that they are not in single file and they are taking up the whole width of the road. Apart from the issues of physical space that John brings up, a big problem is that the current state of the most popular brand of cycling advocacy here in the US (paint and path advocacy) makes it almost a maxim that cyclists should always stick to sidewalks, sidepaths, bike lanes and - if all else fails - the gutter - and then they should ride only in single file (for fear of annoying motorists, most of whom have no idea that passing cyclists riding two-abreast is actually easier and quicker than passing them in single file). If cyclists ride in the way the paint and path advocates say they should, the line of cyclists would stretch back to where the line of cars ends.

So while cyclists choose to deny themselves full and proper use of the road, and while state and local laws continue to try to restrict cyclists' right to the road, cycling can never be a solution.

Another issue that this image fails to illustrate is that the bus is unlike a bike or a car in one fundamental way - it is not a form of personal transportation! Buses are communal, and because they are communal, they do not take a rider from door to door, they are often late, and sometimes they don't run at all. So although it is definitely a space-saver, it is not as efficient (in terms of comfort, speed or ease-of-use) as a bike or a car.

Sunday 5 August 2012

The Lusk Study - A Critical Assessment

In 2011, a study was published online in the BMJ journal 'Injury Prevention'. The study, 'Risk of injury for bicycling on cycle tracks versus in the street' by Anne C. Lusk et al, got a lot of press (as cycling safety studies go), and it is regarded with respect by many in the cycling community. In fact, I think it's probably the single most referenced study of cycling safety in current circulation. But a more critical assessment shows significant flaws.

----------------------------------------

Flaws in the 2010 Lusk Montreal Study


The Lusk study measures bicyclist injury rates on cycle tracks and compares the results with results from supposedly similar reference streets. This was done for six cycle tracks and comparable reference streets. Here I will list the streets that were compared and try to illustrate the flaws I found in terms of the nature of the streets being compared.

1. Rue de Brebeuf Cycle Track vs. Rue St. Denis between Rachel and Laurier.
Rue de Brébeuf Cycle Track
Rue St. Denis

These streets are not comparable.

Brebeuf (which has a cycle track) is a narrow 40kph slow-moving one-way residential street with one traffic lane and one parking lane.

Rue St. Denis (which has no cycle track) is a six-lane (two lanes often taken up by parking) 50kph limit two-way highway in a commercial area with lots of stores and distractions.

It seems to me that more accidents will naturally occur on the six-lane highway with a faster speed limit. It's unsurprising then that the study did indeed find a statistically significant advantage in terms of safety for Rue de Brébeuf. However, I would argue that this has nothing to do with the safety of the cycle track and everything to do with the very different nature of the roads compared. 

2. Rue Rachel Cycle Track vs. Avenue du Mont Royal between St. Urbain and Marquette.
Rue Rachel Cycle Track


Avenue du Mont Royal

Do these streets look at all alike?

Rue Rachel (which has a cycle track) is a four-lane 50kph limit commercial street with a lane apparently open to parking on both sides. On one side, the two-way cycle track is separated from the general traffic lane by a physical barrier and a line of parked cars. Along a third of its length, Rachel has a park next to it, reducing the potential for crossing traffic on this cycle track.

For two thirds of its length, Avenue du Mont Royal (which has no cycle track) is indeed a similar 50kph commercial street with parking both sides. But there the similarity ends. Avenue du Mont Royal has no park and has many intersecting streets along its entire length.

One would think that, if the cycle track was safer, Rue Rachel, with the reduction in crossing conflicts - an unfair advantage - conferred by the park, would show it clearly. Yet the differences in terms of injuries between Rue Rachel and Avenue Mont Royal were statistically insignificant. 

3. Rue Berri Cycle Track vs. Rue St. Denis between Cherrier and Viger.
Rue Berri Cycle Track
St. Denis

These streets are not comparable.

Rue Berri (which has a cycle track) is a 50kph limit divided highway along 1/3rd of its length with the cycle track removed from busy intersections by an underpass, so cyclists are naturally removed from the possibility of intersection accidents.

Along this stretch of Rue St. Denis, the road (which has no cycle track) is a one-way street with a 50kph speed limit. However it is a much busier road than Rue Berri in terms of people doing their business somewhere along that stretch, with a relatively narrow street and lots of intersections and distractions in the form of little shops and cafes along the whole route.

Rue Berri showed a statistically significant reduction in injuries compared with its reference street. However, more accidents are bound to occur where there are lots of intersections and where drivers are likely to be distracted. It seems reasonable that the advantage in terms of reduced injury results on Rue Berri derive from the very different nature of the roads compared and not from the presence of a cycle track.

4. Boulevard de Maisonneuve Cycle Track vs. Rue Sherbrooke and Rue Ste. Catherine between Claremont and Wood.
Blvd de Maisonneuve Bike Track


Blvd de Maisonneuve Bike Track
Rue Sherbrooke
Rue Ste. Catherine

One of these streets is not like the others. Can you spot the differences?

Boulevard de Maisonneuve (which has a cycle track) is a quiet 30kph one-way two lane residential street along much of its length. The bike track goes through a park for 1/5th of its length, thus removing any possibility of intersection conflicts in that area. The presence of the park effectively reduces the chance of traffic collisions by 20%.

Sherbrooke (which has no cycle track) is a downtown 40kph commercial street with four lanes of moving traffic and parking on both sides. It has numerous business distractions along its length. It should be noted that a recent study found that Sherbrooke is the single most dangerous route in Montreal for cyclists. Ste. Catherine (which also has no cycle track) is a similar downtown street, but with a 30kph limit and just two lanes of moving traffic and a lane for parking on both sides.

The idea that these streets are comparable on anything but the most superficial level (i.e. they are streets) is a joke. It is ridiculous, in my view, to attribute a reduction of injuries on Boulevard de Maisonneuve to the presence of a cycle track, when the streets being compared are not at all similar - and when the street with the cycle track has obvious and significant advantages in terms of safety that are unrelated to the bicycle track itself. 

5. Avenue Christophe Columb Cycle Track vs. Rue St. Hubert between Gouin and Jarry and Avenue Christophe Columb between Villeray and Rosemont.

Ave Christophe Columb Cycle Track

St. Hubert
Ave Christophe Columb









These streets do seem comparable.

Avenue Christophe Columb (which has a cycle track) is a 4-5 lane 50kph limit highway along much of its length with one or two extra lanes for parking. The cycle track is completely segregated from the road except at intersections.

Rue St. Hubert appears to be a somewhat quieter 4-5 lane residential street with one or two extra lanes for parking and no cycle track. Avenue Christophe Columb between Villeray and Rosemont (which also has no cycle track) is a 4-5 lane highway along much of its length with a 50kph limit and one or two extra lanes for parking, much like the portion with the cycle track.

I would say this street study would be the most fair in terms of the nature of the streets being compared. So it's interesting that the overall injury rate on these streets showed comparison results that were statistically insignificant. However, it should be noted that in terms of relative danger from vehicular traffic, Avenue Christophe Columb (the street with the cycle tracks) had a much worse (and statistically significant) danger factor. I personally think the way they calculated the danger factor is bogus, but it's interesting that, in this case, it works out to favor the street without the bike path. Also interesting is that the study's authors make no mention of this in their conclusions.

6. Boulevard René Lévesque Cycle Track vs. Rue Sherbrooke between Lorimier and St. Hubert.
Blvd René Lévesque Cycle Track
Rue Sherbrooke

These streets are not comparable.

Boulevard René Lévesque (which has a cycle track) is a very wide six lane divided highway with a 50kph limit and very good visibility along its length. On one side, the two-way cycle track is separated from the general traffic lane by a physical barrier.

Rue Sherbrooke (which has no cycle track) appears to be a slightly narrower undivided 50kph limit road, six lanes wide, with one of the lanes used for parking. The area around the street is somewhat more built-up and closed-in than René Lévesque, so sight lines are not as good. Again, it should be noted that, in choosing Sherbrooke for a reference street, the study authors selected the most dangerous route in Montreal for cyclists.

Rue Sherbrooke has slight disadvantages compared to Boulevard René Lévesque in terms of cyclist safety. However, these are related to the nature of the roads themselves, and not to the presence or absence of bicycle tracks. Again, even though the road with the bicycle track had an unfair advantage, the study found no statistically significant differences in terms of injuries on these roads.

----------------------------------------

Discussion


The study claims:
"Contrary to AASHTO's safety cautions about road-parallel paths and its exclusion of cycle tracks, our results suggest that two-way cycle tracks on one side of the road have either lower or similar injury rates compared with bicycling in the street without bicycle provisions."
Yes, the results do indeed suggest that. But is it any wonder that the Lusk study finds that cycle tracks are safer than street cycling? I don't think so. With the exception of #5, the streets appear to have been chosen in such a way that safety is maximized on the streets with bicycle facilities, while risk is maximized on the so-called 'reference' streets. In my opinion, given the selection of streets used, the study's conclusions could have been foretold before the first cyclist in the study put his foot on a pedal. With the overwhelming nature of the bias in the study, I find it amazing that three of the six street comparisons still showed no safety advantage to cycle tracks!

Here we have what seems to me to be a clear case of selection bias, and selection bias so blatant that I find it hard to believe that the people doing the study did not have a point they wanted to prove at the outset. In fact, the study itself all but admits this agenda, when it almost lapses into rant mode. Note the use of emotive or dismissive language and a very telling admission of a pro- cycle track agenda:
"Cycle track construction has been hampered in the USA by engineering guidance in the American Association of State Highway and Transportation Officials (AASHTO) ‘Guide for the development of bicycle facilities'... A long-standing, and yet not rigorously proved, philosophy in the USA has suggested instead that ‘bicyclists fare best when they behave as, and are treated as, operators of vehicles.’ "
 In my opinion, the integrated (or 'vehicular') cycling philosophy has been rigorously proved (as much as anything can be) by the very studies that Lusk et al go on to dismiss as 'conflicting':
"...crash rate comparisons between the USA and The Netherlands have been dismissed by vehicular cycling proponents, with arguments of non-transferability to the American environment. Cycle tracks have been controversial, especially due to conflicting studies with warnings of increased crash rates."
Contrary to this assertion of conflict, the studies are interesting in that (apart from a very few exceptions) they don't conflict, and that so many of them agree on this specific point: that bike facilities increase the risk of intersection collisions.

Lusk et al go on to say:
"Public health and bicycling advocates in the USA have faced a dichotomy, believing from surveys and European experience that cycle tracks encourage more bicycling, yet being warned that they lead to higher crash and injury rates. Our results suggest that cycle tracks lessen, or at least do not increase, crash and injury rates compared with the street. The construction of cycle tracks should not be discouraged."
So Lusk et al knew that the weight of expert opinion was against bicycle facilities, and they knew that health and bicycling advocates in the US and Canada were finding it hard to move ahead with their objectives while bicycle facilities were known to be hazardous. So it seems they wanted to find a way to redress the 'imbalance'. In short, they seem to have had a pro- bicycle facility agenda.

In an attempt to find other studies that back up their results, Lusk et al attempt to use Wachtel and Lewiston's 1994 study 'Risk factors for bicycle-motor vehicle collisions at intersections', yet they seem to assume that all non-intersection collisions must be on the roadway. Since driveways are not intersections, and since there will be interactions between intersections even for cycle tracks, I'm not sure they can extrapolate so simplistically. Even if they can, they admit that the extrapolated results merely show that separated paths are "at least no more dangerous than bicycling in the street". Hardly a ringing endorsement of cycle tracks. And let's not forget that it was Wachtel and Lewiston who said, in that very same study:
"Bicyclists on a sidewalk or bicycle path incur greater risk than those on the roadway (on average 1.8 times as great), most likely because of blind conflicts at intersections... intersections, construed broadly, are the major point of conflict between bicycles and motor vehicles. Separation of bicycles and motor vehicles leads to blind conflicts at these intersections."
Anne Lusk, judging from the information available on her website at the Harvard School of Public Health, is keen on the health benefits of cycling. I am too. But I don't think misrepresenting data in a way that makes hazards look safer is a good way to help people to become healthy.

Whether the bias was intentional or merely subconscious, the results are the same, and we are talking about people's lives here.

 ----------------------------------------

Implications for Cyclists


At 6:50am on July 24, 2012, a 33 year-old man became the first cyclist (and the first cycle track user) to be killed on any of the streets in the Lusk study, since the study was published. He died on the Christophe Columb cycle track where it intersects Rue Mistral. The driver who killed him may have been dazzled by the morning sun, which would have been in his eyes, and the cyclist may have been partially hidden by the signs on the traffic light post as he approached the intersection. This image shows the driver's view: the cyclist would have been approaching the camera on the cycle track.
The cyclist was killed in a classic intersection turning conflict of the type that bicycle facility experts have been trying to warn cyclists and transportation engineers about for at least four decades. He most likely thought he was perfectly safe - protected by a bicycle track that was physically separated from the roadway. What he almost certainly did not know was that the Lusk study, which supports the facility he was using, admits:
"the American Association of State Highway and Transportation Officials (AASHTO) ‘Guide for the development of bicycle facilities’... cautions against building two-way paths along, but physically separated from, a parallel road. AASHTO states that sidewalk bikeways are unsafe and implies the same about shared-use paths parallel to roads, listing numerous safety concerns and permitting their use only in special situations."
Also, it should be noted that bicycle lane use is no longer mandatory in the province of Quebec. So this cyclist could have legally avoided this bike lane death trap, if he was aware of the change in the law, which happened in 2011. Before then, cyclists were required to use bike lanes.

So far, no roadway cyclists have been killed on any of the study's reference streets. This, despite Quebec's law that forces cyclists to ride on the far right hand side of the road. Presumably this is so that they will be less visible to motorists and therefore more easily killed by them - that way, bike lanes might seem that much less dangerous when everyone finally figures out how hazardous they actually are.

As I said at the beginning, the Lusk study is regarded with respect by many in the cycling community who see it as a confirmation of what we want to hear - that the facilities we expect to keep us safe are indeed as safe as they can be. The problem is, these facilities pose a serious hazard to cyclists, and sadly the Lusk study serves to keep us unaware of that fact. I hope that this assessment serves as a wake-up call to those of us who rely so uncritically on such facilities.

 ----------------------------------------

Since posting this, John Allen alerted me to the fact that there are other critiques of the Lusk study - specifically by M Kary and by Wayne Pein. They both go into far more detail than mine and while mine is merely an amateur critique that addresses only a few of the more obvious (to me) concerns, theirs are far more scholarly and convincing. They can be found here: 

http://john-s-allen.com/reports/montreal-kary.htm

Friday 3 August 2012

Bicycle Infrastructure Studies

Since the idea that bicycle infrastructure increases safety is pretty widely held, I've compiled a list of studies of bicycle infrastructure, including links and what I think are the most interesting quotes. I think many people will be surprised by what the studies suggest:

1972 Deleuw, Cather and Co.: Davis Bicycle Circulation and Safety Study
 http://john-s-allen.com/research/davis_studies/Davis_BicycleCirculationSafetyStudy003_DeLeuw_1972.pdf
"An additional problem is establishment of a visual relationship between motor vehicles and cycles on the sidewalk path on approaches to intersections."

1975 Kaplan: Characteristics of the Regular Adult Bicycle User
http://bikexprt.com/research/kaplan/index.htm
"Surprisingly, bicycle facilities where no motor vehicles are allowed showed the highest accident rate of any variable examined."

1977 Cross: A Study of Bicycle/Motor Vehicle Accidents (USA)
http://ntl.bts.gov/lib/25000/25400/25439/DOT-HS-803-315.pdf
Possible bias in reporting, investigation. Study reaches no clear conclusions about the safety or otherwise of bicycle infrastructure, and many of the conclusions have been called into question by more recent studies. I think the study does remain useful thanks to its detailed crash type analysis.

1987 Grüne Radler review: Police Bicycle Crash Study (Berlin, Germany)
http://john-s-allen.com/research/berlin_1987/index.html
"...with increasing experience, it became ever clearer that the sidepaths are dangerous - more dangerous than riding in the roadway. There is a simple reason for this: the design and location of the sidepaths conflict with the most important principle of traffic safety, the slogan 'Visibility is safety'."

1987 Study, University of Lund (Sweden)
http://www.bikexprt.com/bikepol/facil/sidepath/adfc173.htm#lund
"The basis for the comparison is the crash risk of bicyclists traveling straight through on the roadway. Relative to this, the risk is:
1.1 times for through travel with a bike lane stripe.
3.4 times for a left turn on the roadway
3.4 times for through travel on a sidepath
11.0 times for a left turn from a sidepath
11.9 times for through travel on a sidepath on the left side of the roadway"

1992 Allgemeiner Deutscher Fahrrad-Club: Issues of Bicycling Safety
http://www.bikexprt.com/bikepol/facil/sidepath/adfc173.htm
"Experts from different backgrounds at the Velo Secur traffic safety conference in Salzburg were united in the opinion that sidepaths in urban areas are entirely unsatisfactory in many ways, and should not be used."

1994 Gårder: Safety implications of bicycle paths at signalized intersections (Scandinavia)
http://www.sciencedirect.com/science/article/pii/0001457594900345
"The conclusion that can be drawn so far from combining results shows that the most likely effect of introducing a cycle path is that the risk will increase by about 40% for a passing cyclist."

1994 Wachtel: Risk Factors for Bicycle-Motor Vehicle Collisions at Intersections (Palo Alto, California, USA)
http://www.bicyclinglife.com/Library/Accident-Study.pdf
"Bicyclists on a sidewalk or bicycle path incur greater risk than those on the roadway (on average 1.8 times as great), most likely because of blind conflicts at intersections... intersections, construed broadly, are the major point of conflict between bicycles and motor vehicles. Separation of bicycles and motor vehicles leads to blind conflicts at these intersections."

1997 Moritz: A Survey of North American Bicycle Commuters (USA and Canada)
http://www.bicyclinglife.com/Library/Moritz1.htm
Possible measurement bias: study claims increased safety on bicycle specific infrastructure, but the accident site data appears to be flawed - many of the accidents taking place while on bicycle paths or lanes may have been considered to be on the roadway, because only the final crash site was considered.

1998 Aultman-Hall: Commuter Cyclist On- and Off-Road Incident Rates (Ottawa-Carlton, Canada)
http://www.ncbi.nlm.nih.gov/pubmed/9542542#
"The relative rates for falls and injuries suggest it is safest to cycle on-road followed by off-road paths and trails, and finally least safe on sidewalks... Results suggest a need to discourage sidewalk cycling, and to further investigate the safety of off-road paths/trails."

1998 Moritz: Adult Bicyclists in the United States (USA)
http://www.bicyclinglife.com/Library/Moritz2.htm
"Multi-use trails have a crash rate about 40% greater than would be expected based on the miles cycled on them while cycling on the sidewalk is extremely dangerous."

1998 OECD: Safety of Vulnerable Road Users (European Union)
http://www.oecd.org/sti/transport/roadtransportresearch/2103492.pdf
"The most common conflicting areas between motorised traffic and vulnerable road users are at junctions... While cycle tracks have been found efficient in decreasing bicycle accidents on links, particularly on arterials, they create safety problems at junctions."

1999 Aultman-Hall: Bicycle Commuter Safety Rates (Toronto, Canada)
http://www.sciencedirect.com/science/article/pii/S0001457599000287
"The relative rates for falls and injuries suggest these events are least common on-road followed by off-road paths, and finally most common on sidewalks... These rates suggest a need for detailed analysis of sidewalk and off-road path bicycle safety."

1999 Franklin: Two Decades of the Redway Cycle Paths (Milton Keynes, UK)
http://www.cyclecraft.co.uk/digest/2decades.html
"...the most alarming experience of the Redways is their accident record. Far from realising gains in safety, they have proved over many years to be consistently less safe than even the 'worst case' grid roads for adult cyclists of average competence. This is not an accolade for the grid roads, for their safety performance is not good in relation to lower speed roads of more traditional design. But the segregated Redways have proved to be worse. "

1999 Pasanen: The risks of cycling (Helsinki, Finland)
http://www.bikexprt.com/research/pasanen/helsinki.htm
"At crossings, car drivers focus their attention on other cars rather than on cyclists... the risk of a crossing accident is 3-times higher for cyclists coming from a cycle path than when crossing on the carriageway amongst cars."

2000 Franklin: Cycle Path Safety: A Summary of Research (Worldwide)
http://www.cyclecraft.co.uk/digest/research.html
"little evidence has been found to suggest that cyclists are safer on paths than on roads."

2002 Reid: The Roots of Driver Behaviour Towards Cyclists (UK)
http://abstracts.aetransport.org/paper/index/id/1529/confid/8
"The tendency for drivers to criticise cyclists and to exonerate errors made by drivers can be explained by reference to Social Identity Theory... Drivers regard themselves as intending to behave cautiously around cyclists and yet feel pressurised by other drivers to behave incautiously... It was also notable that drivers rated cyclists as less considerate, even though the cyclist’s behaviour was identical, when encountering them at road narrowings... Cyclists are an ‘out’ group and their behaviour is considered to be inexplicable other than by reference to their status as cyclists."

2007 Jensen: Bicycle Tracks and Lanes, a Before - After Study (Copenhagen, Denmark)
http://trafitec.dk/sites/default/files/publications/bicycle%20tracks%20and%20lanes.pdf
"The safety effects of bicycle tracks in urban areas are an increase of about 10 percent in both crashes and injuries. The safety effects of bicycle lanes in urban areas are an increase of 5 percent in crashes and 15 percent in injuries. Bicyclists’ safety has worsened on roads where bicycle facilities have been implemented."  

2008 Agerholm: Traffic Safety on Bicycle Paths (Western Denmark)
http://vbn.aau.dk/files/14344951/agerholm_et_al._bicycle_paths.pdf
"So the main results are that bicycle paths impair traffic safety and this is mainly due to more accidents at intersections."

2008 Jensen: Road safety and perceived risk of cycle facilities (Copenhagen, Denmark)
http://www.vehicularcyclist.com/copenhagen1.pdf
"The cycle tracks constructed have resulted in increases in accidents and injuries of 9-10% on the reconstructed roads." 

2009 Daniels: Injury crashes with bicyclists at roundabouts (Flanders, Belgium)
http://www.ncbi.nlm.nih.gov/pubmed/19433206
"Regarding all injury crashes with bicyclists, roundabouts with cycle lanes appear to perform significantly worse compared to... other design types" 

2009 Reynolds: The Impact of Transportation Infrastructure on Bicycling Injuries and Crashes: A Review of the Literature
http://www.ncbi.nlm.nih.gov/pubmed/19845962
Cherry picking data: review claims increased safety on bicycle specific infrastructure, but the review cherry picks and misrepresents data - only the 2009 Daniels study (out of 26 studies reviewed) concerned bicycle specific infrastructure safety, and the review misrepresented its findings. 

2011 Lusk: Risk of Injury for Bicycling on Cycle Tracks Versus in the Street (Montreal, Canada)
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3064866/?tool=pubmed
The infamous Lusk study. Selection bias: study claims increased safety on bicycle specific infrastructure, but its street comparisons are flawed - the streets compared were in no way similar other than their general geographic location. Busy downtown streets with multiple distractions per block were twinned with bicycle tracks on quieter roads with fewer intersections and fewer distractions.

2011 Pucher: Bicycling renaissance in North America? (Worldwide)
http://policy.rutgers.edu/faculty/pucher/TRA960_01April2011.pdf
Often cited by infrastructure advocates as a 'bicycle facility safety study', this is a review of studies on cycling trends and policies. It covers safety only in a general sense and while it states an opinion on bicycle facilities, it does not cite any studies pertaining to them. The main point the review makes in terms of cycling safety is in reference to the 'safety in numbers' effect and its ability to increase cycling mode share, but this effect is shown to be false if new cyclists are mostly coming from a much safer mode of transportation, such as mass transit.

2011 Reid: Infrastructure and Cyclist Safety (UK)
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CD8QFjAB&url=http%3A%2F%2Fedoc.difu.de%2Fedoc.php%3Fid%3D31A0Q7OZ&ei=gznBUO7YLpHh0AHHh4HACg&usg=AFQjCNG-d4zNkyS6NLagJo_1T2c3hw2Sxg&cad=rja
"...evidence suggests that the points at which segregated networks intersect with highways offer heightened risk, potentially of sufficient magnitude to offset the safety benefits of removing cyclists from contact with vehicles in other locations." 

2012 Teschke: Route Infrastructure and the Risk of Injuries to Bicyclists: A Case-Crossover Study
http://ajph.aphapublications.org/doi/full/10.2105/AJPH.2012.300762
Selection bias: uses comparison streets instead of a before-after situation; study claims greatly increased safety on cycle tracks, but the cycle tracks chosen for the study were not representative of a typical cycle track, in that all were on roads with limited or nonexistent road intersections. It is not surprising that bicycle facilities that have little or no possibility of interaction with motor vehicles are safer than those that have many such possibilities, and if all bicycle tracks were completely separated from turning and crossing traffic, they would indeed be safer than cycling on the road. The problem is, cycle tracks with few road intersections are very rare indeed. 

2012 Kittleson & Associates Report (Washington DC)
http://ddot.dc.gov/DC/DDOT/Publication%20Files/On%20Your%20Street/Bicycles%20and%20Pedestrians/Bicycles/Bike%20Lanes/DDOT_BicycleFacilityEvaluation_ExecSummary.pdf
Report found:
Bike boxes, bicycle signals and sharrows were installed at the 6 leg intersection of New Hampshire Ave/16th St/U St NW.: after the installation, crashes increased from 4 in 4 years to 5 crashes in 13 months. Per month, that is the equivalent of more than 4 times the number of crashes. The report notes no increase in bicycle volumes.
Pennsylvania center cycletrack: after the installation, crashes increased from 9 in 4 years to 16 crashes in 14 months - 6 times more crashes per month. Taking into account the fact that bicycle volume tripled, crashes still increased by a factor of 2.
15th St NW left side cycletrack: after installation, crashes increased from 20 in 4 years to 13 crashes in 14 months - over twice as many crashes per month. Taking into account the fact that cyclist volumes doubled, this represents an increase in crashes of 10%.
Strangely, despite these significant increases in crashes, the report states that the bicycle facilities "improved conditions for cycling". If this is an improvement, perhaps installing anti-personnel mines every few hundred yards or so might make a bigger 'improvement'.

2012 City of Portland Bureau of Transportation Progress Report: Request to Experiment "9-105(E) - Colored Bike Lanes and Bike Boxes"
https://docs.google.com/file/d/0B-_Kv2GsQnEfNS1pdXRxQkZmcVE/edit?pli=1
"...the crash data trend suggests that right-hook crashes are increasing at some of the treatment locations... We concluded that a high proportion (88%) of the crashes occurred during the ‘stale’ green condition (after the start-up but before the signal phase changes to yellow/red)."
Cycling experts have, for years, been warning about this fundamental flaw in bike box design. The 1997 edition of 'Cyclecraft' by John Franklin advises cyclists that they should approach bike boxes only if the traffic signal is red. If the signal is green, cyclists are advised that the best way to minimize danger may be to stay within the main traffic stream.

That's 28 documents.
21 of which suggest bike facilities are more dangerous than the road.
4 of which suggest bike facilities are safer than the road.
3 of which (1977 Cross, 2002 Reid and 2011 Pucher) do not really address infrastructure safety.

84% of the studies covering bicycle facilities suggest that they are more dangerous than the road. 

John Allen has an extensive set of reports and studies, available here:
http://john-s-allen.com/index.htm
http://bikexprt.com/research/
http://truewheelers.org

John Franklin also has an extensive list, without direct quotes but with his commentary, available here:
http://www.cyclecraft.co.uk/digest/research.html

Bicycling Life also has a set of interesting documents, available here:
http://www.bicyclinglife.com/Library/