Someone sent me some links to other studies of the Idaho Stop. Like the oft-cited Meggs study, both seem to indicate that the Idaho Stop law is either safer or, at least, no less safe.
This study found that crashes at stop-light controlled intersections were more severe in Urbana-Illinois than in Boise, Idaho.
While no significant difference in proportion of crashes exists at intersections verses midblock between the cities, there were differences in the severity of crashes overall. Champaign/Urbana held a 54 percent level of Type B crashes and a 24 percent level of Type C crashes. This compared to Boise’s more balanced 45 percent and 40 percent respectively.
Type B crashes are more severe than Type C crashes.
It is possible that cyclists in Boise are more accustomed to judging the traffic conditions that are safe for crossing. They may be more likely to use caution when crossing a signalized intersection, while their counterparts in Illinois, who are required to stop and wait at all controlled intersections, are not as skilled at judging a safe traffic crossing, thus resulting in a proportionately higher level of severity and one statistically different than in Boise.
Another plausible explanation for these findings is that drivers are more likely to expect cyclists crossing at controlled intersections when the driver has the right of way. Thus, motorists may in fact be more likely to slow down and avoid a crash or greatly decrease the severity of crash.
Bicyclists’ Stopping Behaviors: An Observational Study of Bicyclists’ Patterns and Practices; Catherine Marie Caverly Silva
This is not a study of the Idaho Stop, per se, but it does have a lot of information about it. In addition it studies non-compliance by cyclists at stop signs in Seattle, noting that this is similar to compliance elsewhere.
Results from this study find that approximately 55% of all bicyclists used rolling stops and/or track stands, 25% failing to stop and only 19% coming to a complete stop. Despite this high degree of non-compliance with the stopping law, no reliable evidence was found exhibiting decreased safety resulting from the use of rolling stops by bicyclists at stop signs.
No accidents or near-incidences were observed during the pilot study, which indicates that this significant non-compliance at two-way stops, two-way yields may not be an exceptionally important issue
Other interesting findings
curbside parking, after crossing an intersection, actually inspires more obedient stopping behaviors for bicyclists
a significant relationship between not turning left and not stopping at the stop sign. These findings are not surprising, as it is logical that left turning bicyclists will stop more often, as they need to yield to more lanes of traffic than do bicyclists turning right or traveling straight.
Descriptively and with a consideration for the limited validity of this qualitative variable, these results support the theoretical assertion that bicyclists are capable of making safe decisions regarding rolling stop.
Update: I'd like to see this second study repeated in Boise. Is compliance in Boise different than in Seattle (and New York City, which this study says is similar)? If compliance is the same, it implies that the law and enforcement does little to nothing to change behavior, so the argument for one law or the other is about optics, politics and the "hardship" of enforcement. If compliance is lower in Boise, then that means we need to do more to study whether or not it is safer. If, counter-intuitively, compliance is higher, then we should still study the safety difference but also need to check our assumptions. In all cases, of course, some confounding variable could be the cause of the difference, but that would be more likely in the latter scenario.