Decius wrote:
] ] The report finds that the introduction of cameras
] ] actually interrupted a downward trend in UK accidents.
] ] "If the 1966-93 trend line had continued until 2001 there
] ] would have been 825 fewer fatalities in that year than were
] ] actually recorded."
]
] Cameras do not replace police. Cameras look for speeders.
] Police look for people who aren't paying attention.

... and some people will go looking for the cameras, and those that do probably aren't paying attention, which means that the police should be out there looking for them, which means we haven't really solved anything, have we?

Fortunately for the sake of good return on this investment in infrastructure, the traffic cameras were not installed with the direct intent of "solving" a safety or security problem. To first order, they were installed to raise money. This is consistent with the motivations for traffic cameras in the US.

Traffic cameras do replace or augment police -- the ones who walk the city streets issuing parking tickets, not the ones who drive the streets issuing tickets for speeding or running red lights.

Is the British government "sexing up" the data here? Go back and look carefully at their stated goals. They could still meet their goal of reducing casualties by 40% by 2010, but the path may be rather indirect. The government will collect money from speeding drivers and transfer it to the auto makers through tax incentives to improve crash test performance on future-year models. In absolute terms, high-speed accidents may decrease slightly through investments in guidance, traction control, and other measures. But fatalities and serious injuries due to accidents can be significantly reduced through these and other safety features of future vehicles.

According to the stated goal (as distinct from the claimed rationale), the government does not explicitly seek a direct reduction in the number of road accidents. What they seek is a reduction in fatalities, and this can be achieved through improved vehicle safety. One way to incentivize safety improvements is to offer tax credits to the manufacturers. A fair way to fund those credits is to tax the drivers who are most at risk. You could argue about whether the threshold should be 10 miles over or 15 miles over, but this is a detail. (A fee schedule with a ramp-up can solve this problem.) If 42 cameras across one large city can raise $100M per year, then 8400 cameras across the world's two hundred most populous cities could raise $20B annually. By the 2010 model year, we could see the fruits of a $140 billion investment in vehicle safety. It is worth noting that the desired effect will be delayed relative to the installation of cameras due to the indirect path of the funding and the time required for drivers to buy new vehicles. It would be unrealistic to expect an immediate response.

The author of the study seems to be a statistician making a rant about alleged sloppy data processing. Interestingly, he claims to have evidence against the popular notion that commercial airlines are the safest form of transportation.

Also, the author repeats the well-known idea that "correlation does not imply causation", and then he whines for a while. Later, he points to a recent absence of a drop (but not an actual uptick) in fatalities and then blames it on cameras. Case in point.

The "Christmas Double Demerit Period" is an attempt to further increase revenues by capitalizing on the holiday driving season. People are out shopping, traveling to visit relatives, and so forth. It also plays well in the news, because no one likes to hear about driving deaths during the holidays, and this seems like such a caring, well-intended measure.

With regard to this study, the best outcome would be for the government to explain the traffic camera tickets as an excise tax rather than as a punishment. A driver should not feel any more shame when paying one of these tickets than she does when paying the annual registration fee on her Mercedes SUV.

RE: [Politech] New study says speed cameras don't work [priv]