The point about “when people are expected to break the law” is strange. You are never expected to steal things in normal life, and if you accidentally take something, you give it back right away, with the acknowledgement of fault. In cases like manslaughter, where we do expect fatalities in the course of operations, a lot of the penalty has nothing to do with changing or deterring specific “normal” behavior, but rather ensuring that people can’t commit murder with impunity via hitting bicyclists with cars, etc.
The point about continuous enforcement is also strange. In non-normal life, such as natural disasters, people will loot (or irregularly buy, by placing contact info and cash on empty counters, etc.) stores, but that occurs in a context of people being so pressed for survival that continuous enforcement is not viable. Indeed, the concept itself is strange: if you have the capability for continuous enforcement, why would you ever give people the option to break the law in the first place? What resource setting would put you into such a position, where you had the capability to enforce the law, but not to prevent or to deter people from breaking that law?
In an properly ruled liberal society (not a gangster tax situation where you lay a trap), I can’t think of any other example than speed limits where ordinary people, in an ordinary situation, are frequently enough, put into a position where they have to choose between safe behavior and obeying the law. (But I would be very interested to know/be reminded about other such situations.)
The basic problem is:
- Maintaining an exact speed limit is less safe than varying it in response to conditions
and additionally, the three common mitigations or responses to continuous enforcement also increase danger:
- Maintaining a significantly lower cruising speed (that would allow variation in a band below the enforced limit) than the posted limit increases the risk of rear-end collisions, sideswipes, and pileups; not-to-exceed speed driving creates similar risks
- Use of cruise control, in situations that aren’t driving on relatively straight roads in fair enough weather with well-spaced out traffic, is significantly more dangerous for a wide variety of context-dependent reasons, than manual operation
- Frequently checking the speedometer or other control systems or maps (e.g. to check the correct speed limit for the area), excessively diverts the driver’s attention away from the road and surroundings, to the point that it constitutes distracted driving.
and the combination of these dangers, when applied to areas with significant speed (e.g. above 25 MPH) may very well be greater than the safety benefit of an overall speed reduction, particularly when adequate intermittent enforcement is deployed to cut down on reckless driving.
(a minor point but worth noting) Moreover, flawed continuous enforcement e.g. due to legitimate attempts to maintain normal driving expectations while signaling to drivers the appropriate speed, rewards drivers for driving faster/with less situational awareness, and for driving on roads where compliance/cost is reduced, instead of the roads best intended/suitable for the traffic.
Statistically, the benefit of said continuous enforcement is uncertain, and so it might be that continuous speed limit enforcement has disproven the essential danger of mild speeding.
The detailed explanation follows.
The primary safety purpose of traffic control regulation is to prevent collisions. Collisions occur only when two entities occupy the same space; therefore, traffic control regulations must be directing drivers to take measures that provide additional spacing (e.g. traffic light), or that enable significant increases in the amount of practical capability to avoid a collision (e.g. posting a speed limit in a blind corner that will ensure a driver who sees a slow bicyclist has enough time to brake and/or dodge).
The secondary safety purpose is to provide guidance to burdened/less capable road users like tractor-trailers, landscapers towing equipment, etc. about the safe speed of the road, to include things like road camber, curves and humps, etc. when they likely can’t predict how fast they should be going. The major caveat is that e.g. modern sports cars on high performance tires, are not limited by control capability, but by visibility.
The tertiary safety purpose is to limit the damage of collisions; this could be by setting a limit low enough to allow a pedestrian to survive a collision, or for a head-on car crash to allow both parties to walk away.
(Other purposes of traffic control regulation include conservation of fuel and limitation of damage to the road e.g. by heavy trailers, overload of bridges.)
Consequently, proper safety applications of the speed limit are:
- Conveying the maximum safe speed to burdened road users
- In pedestrian-heavy zones, a very low speed limit may reduce fatalities (but in many situations, this is hard or impossible statistically to demonstrate; for example, in the center of London or New York City, you might not be able to drive faster due to congestion; moreover, a collision may involve rolling over the vulnerable party, knocking them into other traffic. Likewise, if all the pedestrian fatalities are caused by reckless driving, which is faintly influenced by modifying the speed limit, you likely would not see a change in fatalities.)
- By the similar thinking, on undivided roads especially, a very low limit (25 miles an hour) could theoretically eliminate the loss of life by vehicle occupants
- To set the expectation of which users are on the road, and how fast they will be going (e.g. highway minimum speeds)
and to that last point, people put their hazards on, yellow tape, etc. if they are operating a vehicle significantly below said expectation, as it is recognized that under such conditions, a slow-moving vehicle is a hazard.
However, keeping to the speed expectation (to say nothing of keeping with the actual flow of traffic), is not sufficient to lower collision rates to their practical minimum. Cars have to move out of blind spots (particularly trucks), and also must facilitate effective merging in and out of traffic. This requires cars to move lanes, decrease or increase speed – the latter two of which are not complying with the number of the speed limit, and for which the first one may also indicate a significant increase in speed. On American freeways, safe speed increases and decreases for these purposes, routinely exceed 10 MPH. On American back roads, such as in the mountains where you do not want lines of traffic tailgating/tempting rear-end collisions, and making it difficult for opposing traffic to share the road, a practical speed increase can easily exceed 20 MPH, due to the short visibility and time window to make the pass, especially for drivers not familiar with the road. Hence, to obey the basic safety principles of maintaining adequate separation of traffic, but roughly meeting the expectation of how fast traffic will be going, the speed limit has to be disobeyed.
There are other situations that call for contravening these limits. One such special case is hills and mountains, especially for larger vehicles. To avoid grossly unsafe brake fade, you coast down the hill and let the uphill bleed off the speed. Even a relatively small decline or incline can cause 10 MPH fluctuation without additional (or any) throttle input. Another case is rain, fog, and moisture, where although the vehicle mostly would be physically capable of maintaining the speed (possibly unlike snow/ice), you would not want drivers to be forcing themselves to maintain the speed limit if visibility or car control became challenging as a result. Likewise, if a lot of cross traffic, pedestrian movements, etc. started happening adjacent to the roadway, you would expect drivers to slow down while this was occurring, and then speed up after the road area became less crowded.
Hence, in the operation of a motor vehicle under normally varying American road conditions, maintaining an exact speed throughout a trip, is less safe than varying speed up and down.
Thus, one major issue with drivers slowing down only to avoid enforcement, is the decrease in vehicle distance caused by violated speed expectations. Some drivers expect traffic to go at the posted speed, and people trying to ensure they avoid enforcement go at a speed that ensures their variation remains under posted. This causes vehicles to approach each other, so the approaching driver has to take action to avoid a collision. Thus, this behavior creates more hazards relative to a looser approach to speed determination that goes with the flow of traffic. When otherwise capable road users (not just the burdened road users who have other reasons to do so) are driving at the slower speeds, this increases the number of hazards, up until the point that most road users are driving at a significantly lower speed. In theory, this could work along the lines of “everyone normally goes 10 MPH under” but usually, the margin for error is not officially or informally communicated, so this adjustment doesn’t happen. Also, in practice, given e.g. a speed limit of 40 MPH (therefore 10 MPH under is 30 MPH), this means law-abiding traffic can be 25% or more slower, which is a large enough penalty that more affluent drivers will consider enforcement as a fee for faster travel, like a toll road.
The second major and related issue, is the prolonging of inadequately spaced vehicles. If a driver is governed (physically or by e.g. not wanting to get a ticket) not to exceed a certain speed, that driver is not going to speed up to re-establish adequate spacing in all directions. For example, drivers almost universally are familiar with people blocking both of two highway lanes. This is an unsafe practice because any sideways deviation by either of the two vehicles, will cause a collision, and the train of vehicles behind them will then cause a gross pileup. As noted above, similar issues occur in merging situations.
Anything that causes vehicles to get close to things is going to decrease safety.
Cruise control usage is, at a high level, less safe than a half-attentive driver. The vehicle is set on an automatic course to run into something; moreover, non-autonomous cars don’t have enough capability to handle common situations like merging. Primary situations where cruise control is helpful are:
- Reducing driver fatigue on long, but uncrowded, trips
- As a safety check when making passes, and the driver’s eyes and attention are focused away from the immediate traffic ahead (looking down the road, over shoulder), adaptive cruise control can help reduce rear-end crashes
and the primary road type on which it is feasible is a long, straight, stretch of road, with few entry points. In other words, cruise control is most helpful and practical on the roads which need the fewest traffic control regulations (including speed limit enforcement).
Even in nominal, sunny-day or calm night cases, adaptive cruise control does not uniformly improve safety; in addition to all issues about driver vs. vehicle control (does the driver realize whether the system is operating), the system may slow down in response to phantom phenomena, or speed up when it can’t recognize obstacles ahead. This may cause significant acceleration during lane changes and other situations like exiting a road where continuing the same speed may be safer; this is why you should consider disengaging the system when the situation gets more complicated.
Cruise control is unsafe under adverse road conditions. If the unaware system speeds up into standing water, control can be lost; letting it maintain speed through wet corners is testing the tires’ traction; and typical existing systems do not slow down ahead of blind corners. In the case of road obstacles, such as speed humps or other traffic calming, cruise control at typically posted speeds (higher than the traction limit of the traffic calming) will cause a loss of vehicle control. Moreover, existing systems deployed on standard vehicles (not $100K+ Waymos with 20+ sensors) do not try to anticipate entry or exit from the roadway e.g. by animals or pedestrians. Hence, use of typical cruise control in highly urban areas, school zones, walkable pedestrian areas, hilly and tight roads, etc. increases danger to road users, especially the most vulnerable ones.
Cruise control also becomes less feasible as the number of traffic control regulations increase. If a road has many intersections, to use this function, the driver frequently disengages and re-engages the system. The similar problem occurs when speed limits vary, especially on the same stretch of road. Once the driver is readjusting the system every minute or less, it’s no longer automatic speed governance; consequently, the “just use cruise control” is not a constructive suggestion for how to avoid continuous enforcement actions.
Due to these above considerations, cruise control is a particularly infeasible approach to governing speed in school zones and other high-pedestrian or other obstacle-prone sections where specific traffic enforcement typically is deployed. The highest level of safety in these situations is achieved when drivers focus on governing their speed based on what isn’t in the road (vs. on the freeway when you want the drivers mostly paying attention to things in their side of the freeway). As such, having heavy enforcement too far ahead of the buffer to the danger zone (that is, you want to give drivers time to adjust speed/approach before they enter the zone) causes drivers to have cruise control operating at the very moment in which it is most dangerous.
Frequently looking down to check speed is an obviously dangerous driving practice. What is not necessarily so evident, is how many situations can encourage this behavior:
- Speed limits that are far out of line with the road configuration. When you see 35 MPH signs for asphalt area larger than 55 MPH rural highways, and the road itself has few features such as intersecting roads, you have to look down and check.
- Speed limits that are inconsistently set among neighboring roads. When a 25 MPH road with great sightlines in most of the stretch, transitions to a 40MPH two-lane road with many narrow entrances with no road shoulder, what a driver “feels”, or is actually, safe, is in no way correlated with the posted limits.
- Posting initial enforcement signs (or any signs) very close to the area of heightened enforcement, so that at the exact moment when the driver should be looking e.g. for schoolchildren, the driver’s eyes are now on the gauge cluster.
- Frequent stop signs, traffic calming, etc. where the driver has to reduce or stop, and re-gain speed.
(A minor point) In certain situations, continuous enforcement protocols reward drivers for disobeying speed limit signs, that notionally, would help reduce e.g. pedestrian collisions. For example, if you have a stretch of roads/paths that vary between 30 and 40 MPH posted speed limits, and your continuous enforcement threshold is at 8 MPH over (notionally reasonable variation), then a driver will avoid getting pulled over/fines, by setting cruise control to 35 MPH throughout this stretch of roadway. Thus, the continuous enforcement regime is not merely undoing the signaling function of the roadway and causing drivers to drive faster and with less regard for conditions; it is making the violation of expectations worse by giving contradictory signals e.g. to drivers that don’t normally drive the road / realize the continuous enforcement is in place, and those that do / are regulars.
Said behavior generalizes to further undesirable behavior, both safety-related and purely economic. If the road conditions, changes in posted limits, camera intervals, etc. make driving a particular set of roads more costly / effectively tolled due to continuous enforcement, then drivers will choose their routes based on the effective toll cost and not on other factors such as overall trip times, is it a divided highway vs. a two-lane road, residential vs. planned boulevard/stroad/local artery. Depending on local conditions, this can lead to:
- People choosing to live/work/play/dine in different areas partially based on the effective toll
- Increase in residential road traffic overall (maybe not safer and definitely noisier)
- Driving people to the same roads where the effective toll rate can be minimized by maintaining constant speed, possibly overloading the road by using the road for longer-distance travel for which it was not intended
and a further contributor to this type of issue is the extent of any mileage-based tolling on other roads. For example, if a highway notionally would be the preferred route for longer-distance travel, but that highway is tolled, then even more traffic gets steered onto the lowest-cost (likely a less divided and more traffic controlled) route vs. the lowest-travel-time, safest, etc.
Finally, we should consider whether the net decrease in speed in areas of heavy enforcement, typically outweighs the danger of overall more dangerous drivers. Changes (positive and negative) in driver behavior necessarily diffuse over all road areas, so it is fundamentally hard to isolate the effects. Moreover, drivers of all kinds are going to avoid continuously enforced areas, so there is a traffic volume reduction relative to the baseline that should be expected, and so the volume reduction (not necessarily characterized) may be confounding improvements. Moreover, net decrease in speed is one factor, but others (such as red light cameras, pedestrian improvements, etc.) also play.
I don’t think the question easily is answerable via statistics, although we should try. First, we should orient ourselves to as to what statistical effect we could observe if we tried to alter well-understood dangerous driving practices; for example, bans on handheld cell phone usage and texting.
This is a well studied-topic; reports range from 2% to 7% reduction in serious injuries including death, and maybe even up to double-digit improvements in fatalities only. However, primarily, given the wide reported ranges, I remain cautious about using this as an anchor point.
The statistical situation does not seem to be better in regards to the effects of speed cameras and modifications in continuous enforcement. In reviewing claims of effectiveness and data such as the following:
https://www.iihs.org/topics/bibliography/ref/2097
** (also note that Montgomery County, Maryland’s population increased over this time period, while vehicle miles traveled stayed roughly flat for a decade before 2015, when it spiked upward (measurement artifact?)) Note also, I discuss the data from 2020 backward, since the reset in 2020 changed the baseline significantly. (looking at some data from 2020 onward, there are a lot of not obvious features when combined with this data)
there are a lot of difficulties in determining whether the speed camera interventions (which are confounded by other improvements and changes) improved the situation with regards to serious injuries, fatalities, and vulnerable population collisions. The intermediate measures of effectiveness cited in the IIHS review of data, suggest drastic decreases in average speed, and extreme awareness amongst the driving population of the changes. That is, if the hypothesis is that increased enforcement will reduce average and peak speeds, and reductions in these speeds will decrease safety-related adverse incidents, the reported data indicated that speeds clearly decreased.
However, in reviewing the most favorable indication trend line that shows the data pre-intervention (fatal crashes per year, vehicle occupants), the claim that 2012 changes to a “speed corridor” system improved the situation, is not at all obvious: 3 of the preceding 4 years had fewer fatalities than the 3 years after the 2012 changes. The pedestrian data points have similar issues. The effect looks stronger when considering the 2007 initial implementation, but it’s not that obvious. Neither of those “new” baselines align well with the decreases in 2016+ period after the study concluded; this is especially important because vehicle miles traveled increased (which, all else equal, should have increased deaths and injuries). Since the ordinary effect of vehicle miles traveled is not obviously reflected in the data, this leads us to have much lower confidence in our understanding of the overall situation. Considering serious and fatal accidents in the chart above, one might hypothesize that overall, drivers were less reckless, but then when you look at the pedestrians on that very same graph, or on the overall collision data, you see essentially no change in the same time period.
Several things stand out in trying to make sense of all of this:
- Inadequately long baseline period to measure effects in this easy to find data
- Gross doubts about whether all of the relevant variables were collected to prove any hypothesis
- Small number problems, especially for pedestrians and fatalities
- Graphs that indicate effects or flat lines, but that don’t line up with each other to tell the same story. In other words, these data seem very prone to cherry-picking, for and against the hypothesis that the speed cameras made Montgomery County roads safer.
The part that troubles me the most about all of this, is that if we credit the intermediate MOE of speed reduction and awareness, that the most obvious interpretation of this data is even if you lower effective speed by 10 percent and lowering “excessive” speed by even more, you only will see a decrease in motor vehicle occupant serious injuries; you will see at best (and as noted above, very questionable whether that’s even the right directionality) a minor decrease in fatalities. Assuming we were to credit (probably incorrectly in attribution to start with, the trend line is smooth, that suggests cars got safer, not speed – a longer baseline would help) the decrease in serious injuries to speed reduction, that means the speed reduction only operated in a range that prevented serious injuries (which, combined with the IIHS study, could suggest going from say, 50 MPH to 35 MPH), but could not reduce the death rates. That narrow way of thinking is not physically consistent given that the 50 MPH range is above the ~43 MPH at which vehicle passive safety becomes ineffective; fatalities should have been converted into serious injuries at some rate, and the trend lines of those things do not line up. It also would suggest that there is a high probability that those fatalities and the overall collision rates cannot effectively be reduced by any reasonable amount of additional enforcement under the current system of road operation (e.g. not 25 MPH everywhere). Statistically, perhaps one could accept that the speed cameras changed serious injuries to walk-aways, but, if we are in the business of crediting the apparent effects of the 2012 changes, the data suggest that those changes killed extra people. Is it a worthwhile tradeoff to get 100 fewer serious accidents in exchange for 5-6 more dead people a year? Is that “safer”?
An attempt at sanity checking: if we credited the cellphone policy changes with 2-7% change in serious injuries including death, that’s a difficult statistical signal to pull out of data series like these. While we try to let the data guide us, if we’re coming into this with a bias that we should see an effect like this, to see data series that alternately decrease by more than this, go flat line, show punctuated equilibrium, etc. – it is not obvious that this set of continuous enforcement policies is statistically similar, which means these two sets of policies for this jurisdiction probably aren’t similar in their effects on safety.
In summary, none of the statistics I have reviewed on these points are heartwarming. When statistics are not helping, you abide by sound principles and “what do I want law-abiding people to do”. Or, if it seems like a wash, because you can’t figure out which of these principles would have more effect, you look at other things, like cost and attractiveness/do people want to live, work, shop, visit, under that regime.