Elizabeth Macdonald, Rebecca Sanders, and Alia Anderson write: Performance Measures for Complete, Green Streets: A Proposal for Urban Arterials in California:

"In the State of California, speed limits are set using requirements in the California Vehicle Code, which states that the speed on multilane State highways (which includes State urban arterials) will be 55 MPH unless a traffic and engineering study has shown that speed is not reasonable or safe in that location. On the other hand, on a non-State highway in a business or residential district, the Vehicle Code sets the speed limit at 25 MPH. Although these speeds can be adjusted by the DOT or by the local government through a series of studies and petitions, it does not seem reasonable that, in urban areas, State and local arterials should be treated so differently.

Furthermore, localities can petition to have their speed changed if they demonstrate that 85% of drivers are driving a certain speed. In other words, the 85th percentile rule adjusts the law (speed limit) to fit the behavior (actual speed). According to the Vehicle Code, “a reasonable speed limit is one that conforms to the actual behavior of the majority of motorists, and by measuring motorists’ speeds, one will be able to select a speed limit that is both reasonable and effective.” While this system may be appropriate on freeways and major highways, it is not suited to urban environments where roads are shared by a variety of users. Research has shown that posted speed limit signs appear to have a limited effect on reducing driver speeds when not accompanied by enforcement and roadway design.108 While enforcement can be effective, it is a reactive approach that is limited by financial resources. The most proactive and long-term approach is to design arterials for the safest and most appropriate behavior (actual speed) for each location."

I have long doubted the reasonableness of the 85th Percentile Rule. Why 85th Percentile? Why any percentile? Presumably it has to do with safety. Copenhagenize discusses this today, and suggests it is due to the Solomon curve. I think it is older than that (1964), but have not been able to find the source. [Google Ngram viewer suggests 1959 as the first mention of the term 85th percentile speed, but that does not tell us about speed limit rules, which I don't see until 1981, and that is far too late, maybe someone has a better query.]

Solomon's curve is not gospel, Davis et al. (2006) Speed as a risk factor in serious run-off-road crashes: Bayesian Case-Control Analysis with Case Speed Uncertainty could not corroborate it, finding higher speeds associated with higher likelihood of certain crash types, but not lower speeds.

We can think of the rule as potentially acting as a positive feedback system, an upward moving ratchet for speed limits. In year one, speed is set at 30 mph but many drivers exceed it and the 85th percentile speed is found to be 45 mph. The new speed limit is set at 45 mph. Before 15% of drivers were exceeding 45 MPH, but now some drivers, seeing the higher speed limit, drive faster. (I.e. following +10 mph rule, that you won't get ticketed if you are within 10 mph of the speed limit, you get more speeders). So the new 85th percentile speed is higher still. Sure there is an upper limit to this (e.g. if the speed limit were 100 mph, we would not get 15% of drivers exceeding it), but as noted in Copenhagenize what is safe for drivers may not be safe for other road users (especially pedestrian, bicyclists, neighbors).

There are lots of alternative strategies for setting speed limits, and rules for urban areas should differ from rural areas. Perhaps we don't need explicit speed limits, which act as both a ceiling and a floor, everywhere if instead we moved toward self-explaining roads and shared space. Raising speed limits on freeways may improve overall safety (e.g. Lave and Elias) if you take impatient traffic off of non-freeways. The issues are complex.