Data is good, as is statistical analysis of that data. The cited studies leading some to conclude one should "never" cycle on a sidewalk, however, are underwhelming.
The first,
http://www.bicyclinglife.com/Library/Accident-Study.pdf, studies non-fatality accidents in a
single city (Palo Alto), and defines the overall population of cyclists by observations during a
single 8 hour period. While this is not necessarily a case of GIGO, it isn't exactly a broad based study.
The second,
http://www.bicyclinglife.com/Library/Moritz2.htm seems to have a decent population (League of American Bicyclist members - presumably good riders), so that could be meaningful. The survey results indicate a higher frequency of accidents per miles traveled on "other" facilities: "An 'Other' category appears in the present study and nearly all responses indicated this meant sidewalks. They accounted for just 0.3% of the kilometers ridden." But (because these are check-the-box survey results?) there is no analysis beyond calculating numerical ratios. E.g., for the purposes of this thread, it would be helpful to understand why the cyclist was on the sidewalk, cause of the accident, etc.
The first study starts with:
"In 1992, 722 bicyclists were killed in the United States in collisions with motor vehicles, and an estimated 650,000 people were treated in emergency rooms for bicycle-related injuries. It is remarkable that, for a traffic safety problem of this magnitude, so little research has been conducted to establish the causes of these accidents."
As one is often more interested in what might be fatal vs. what might cause a scraped knee, the second study had a reference,
http://www-nrd.nhtsa.dot.gov/Pubs/FARS94.pdf, that contains the table below. Unfortunately it does not state if any did or did not occur due to sidewalk use.
Table 99
Pedalcyclists Killed, by Related Factors
Factors | Number | Percent |
Failure to yield right of way | 203 | 25.3 |
Riding, playing, working, etc., in roadway | 123 | 15.3 |
Improper crossing of roadway or intersection | 101 | 12.6 |
Failure to obey (e.g., signs, control devices, officers) | 75 | 9.4 |
Inattentive (talking, eating, etc.) | 48 | 6 |
Failure to keep in proper lane or running off road | 46 | 5.7 |
Erratic, reckless, careless, or negligent operation | 32 | 4 |
Operating without required equipment | 30 | 3.7 |
Making improper turn | 26 | 3.2 |
Driving on wrong side of road | 22 | 2.7 |
Improper lane changing | 21 | 2.6 |
Not visible | 20 | 2.5 |
Improper entry to or exit from trafficway | 16 | 2 |
Failing to have lights on when required | 12 | 1.5 |
Other factors | 75 | 9.4 |
None reported | 220 | 27.4 |
Unknown | 24 | 3 |
Total | 802 | 100 |
Note: The sum of the numbers and percentages is greater than total pedalcyclists killed as more than one factor may be present for the same pedalcyclist.
A good compilation of studies is
http://www.bikexprt.com/bikepol/facil/sidepath/sidecrash.htm. It includes graphs based on the Moritz study mentioned above, however turning not-well-explained tables into pretty graphs does not increase the information quality. There does seem to be consistency in the results from various studies, finding a positive correlation between accidents and sidewalk use, but the quality and quantity of underlying data remains underwhelming.
Personally I think it has more to do with poor cycling practice on sidewalks (too fast; not considering that cars aren't expecting a cyclist to be there; etc.), and not that any road is safer for a cyclist than its adjoining sidewalk.
And one final comment from
http://www.annarbor.com/news/ann-arbor-mayor-john-hieftje-kicking-around-idea-of-banning-bicycles-on-downtown-sidewalks/"BBC reported zero pedestrians killed by cyclists in 2009, but 426 were killed by cars. There were 13,272 collisions between bikes and cars in 2008, 52 of which resulted in a fatality for the cyclist. No drivers were killed."
In sum, there are times and places to use the road, and times and places to not use the road.
Caveat cyclist in any case.