A mathematical analysis that I'd always assumed was the case, but had never actually seen. This is where smarter cars could help.

Not being cynical here, but I thought industrial and process engineers have known for a long time the negative effect of variance on a system. I've told my kids for years that the guy weaving in and out of traffic to get ahead quickly is just as much a cause of traffic jams is the guy driving too slow in the left hand lane. Indeed, this has always been my problem with those who tailgate, because they have to react more than the person in front of them. To compensate, the person following the tailgater has to keep a greater difference to allow reaction time to determine if a brake light is a slight adjustment or a hard brake.

The research team now plans to develop a model for cars equipped with new electronic devices, which could cut down on over-braking as a result of slow reactions.

This is easy (now being cynical). They called it governors previously. I'm not sure people will like it, but if you limit how quickly cars can vary speed, then you reduced the problem. Next they can put up speed limits to encourage drivers to maintain a standard velocity. Eventually, they might try cruise control with look ahead radar to adjust speed for slight variances in standard velocity...

My observation is that many traffic jams are caused or exacerbated by police officers who leave their lights flashing even when they are parked away from the road shoulder. Also a lot of drivers are idiots -- e.g., people who slow to look at an accident scene are not only creating a problem for other drivers but also, by distracting themselves, reducing their own ability to respond to other drivers' idiotic behavior.

This is no surprise to me. I've always been aware that stop-and-go driving habits cause and worsen bad traffic. The way to avoid this is to leave more space ahead of your car and accelerate or brake more gradually, when there's traffic. Try to drive at a constant speed as much as possible. By driving more regularly you tend to cause the cars behind you to drive more regularly as well. Also, by leaving more space ahead of your car you make it possible for cars to merge into your lane smoothly without requiring you to stop or slow. I erase clumps of traffic jams all the time when I drive, all it takes is a little patience.

Pretty dang obvious; I suppose there must be some benefit to having it modeled accurately, though.

My observation is that many traffic jams are caused or exacerbated by police officers who leave their lights flashing even when they are parked away from the road shoulder. Also a lot of drivers are idiots -- e.g., people who slow to look at an accident scene are not only creating a problem for other drivers but also, by distracting themselves, reducing their own ability to respond to other drivers' idiotic behavior.

The rubbernecking problem is overrated. I think vehicles such as police cars that are routinely parked on the side of the road should be highly visible. Second, you shouldn't be speeding past parked vehicles or accidents.

The rubbernecking problem is overrated.
I don't think so; it's common here in NorCal to have traffic backups in the lanes going the opposite direction from the lanes with a traffic accident. and the backup goes away almost immediately thereafter.

The complement to not waiting to brake is accelerating slowly after a jam appears to clear. Yes, after 10 minutes of stop and go it's nice to get back up to 65, but typically that lasts less than a mile, and one is back down to 25 again. Accelerate slowly to 50 and you don't catch the pack in front of you so quickly (but you -do- have a lot of people passing you, building up the pack at the back and bringing the back edge closer to you.)

I thought this was old hat. I remember reading what sounded like the same theory/analysis many years ago. Comments on the linked page seem to make the same point.

On I-95, northbound between Richmond and Fredericksburg, there is a big sign that spans the three lanes. Guess where the multi-mile traffic back-up starts? People slowing to read the sign that warns of heavy traffic 20 miles ahead. Unbelievable.

I think vehicles such as police cars that are routinely parked on the side of the road should be highly visible.

As another commenter pointed out, drivers often slow in response to distractions in the opposite traffic lane. I think it would help if police took this behavior into account when deciding whether to leave their flashing lights on. I have many times been in daytime traffic jams, when weather and visibility were excellent, that evaporated as soon as the pack passed a parked police car with its flashers on. Sometimes these police cars are parked well off the road and have no need of flashers. I think that what happens in some of these cases is that there is an accident to which the police attend, and then they leave their flashing lights on long after the accident scene has been cleared and their police vehicle is the only car remaining.

My theory out to be testable.

Perhaps the reporting was bad, but I remember learning this in an undergraduate applied math class... um... close to two decades ago.

Other interesting math includes traffic lights (they're disasterous, for much the same reason, though correct timings between them can somewhat limit the damage), and how minimal a safety margin people allow in terms of reaction time (any engineer who designed a system with such a minimal margin of safety in the control systems would be fired - it's a wonder drunk drivers ever make it home at all).

Perhaps the reporting was bad, but I remember learning this in an undergraduate applied math class... um... close to two decades ago.

There is no "perhaps" about it. The scientists undoubtedly found some new ideas on the theme of traffic simulation. The article conflated background with new results.

Seattle scientist and Uber-Geek Bill Beaty has this up on his website with a 1998 publishing date. Observation, analysis and a cure.

http://www.amasci.com/traffic