Your policy conclusions may have some validity. Your statistical arguments have none.

Automobile accidents are widely distributed, have numerous causes, are largely nonintentional, and, more to the point, have likely remained more or less constant before and after Sept. 11. There may be small observable trends, but the gross picture hasn't changed.

You could have predicted in January, 2001, that there would be about 3,000 - 3,500 deaths due to traffic accidents in September.

The same cannot be said for terrorism.

Terrorist acts are not widely distributed, have a single cause (intention of the individuals or organizations perpetuating them), are intentional, and, more to the point, have changed rather drastically in perception (at least to American minds, and I'd posit most of the world) since Sept. 11, 2001.

As your friends in insurance will tell you, terrorism isn't insurable, in the ordinary sense of the word. You can buy terrorism insurance, at a very steep premium. But, due to the intentional nature of terrorism, the very act of doing so may change your odds of being targeted. By contrast, you can buy health insurance with suicide coverage, and it's a fairly standard rider -- but the boys uptown won't let it cash out until you've carried the policy a few years (two, IIRC). Their numbers show that this is a predictable risk (few suicides practice sufficient self restraint to, er, profit).

In an environment with sufficient sustained medium-level incidents -- say, Israel or Sri Lanka -- there might be grounds to provide a moderate, capped, general payout for terrorism, so long as trends don't change.

The problem, however, is that statistics -- the use of averages, means, trends, and laws of large numbers -- to predict intentional events of large scale, is misguided. The assumption in statistics is that past experience is a guide for the future, that there are corralative (and possibly causal) relationships which can be drawn, and that for a given set of measures (with a given cost), outcomes may be swayed.

In your automobile example, it's reasonable to assume that, to a good approximation, countermeasures to automobile deaths were largely at appropriate levels prior to Sept. 11, 2001: there's not much a person could be doing differently today than she would have done September 10. Wear a seatbelt, drive a car with an airbag, avoid drugs and alcohol, obey traffic controls, stick to well-engineered routes, avoid driving at night, in inclement weather, or 2am on weekend nights.

Terrorism violates all of those assumptions.

To a gross level, there are some elements of probability which can be suggested. Population, government, business, cultural, military, and infrastructure centers are more likely targets. Surprise is certainly a tool of the terrorist. Sustained levels of global defense are not possible -- there are too many targets, and they cannot all be guarded, all the time. IMO there are many lessons to be learned from computer security: diversify your resources, provide for redundant capabilities, back up essential resources, defend against obvious threats, watch for signs of hostile activity, and take early action to minimize the damage resulting from an attack. Note too that the enemy's actions have all leveraged grossly inferior capabilities, largly by hijacking our own infrastructure.

Note that the last doesn't mean "hunt down and destroy the perps of every offense" -- particular cells may not be worthwhile targets (hitting the enemy's nervous system is certainly an objective). It means remove their ability to harm you. This can be by disarming, disabling, immobilizing, or destroying the enemy. It can also be by removing or reducing the risk aspects of their target. Move sensitive populations or materiel from the site. Apply countermeasures (drugs, immunizations, fire supressants) to the particular threat mode. Remove volatile loads (fuel, explosives, chemicals, etc.) or render them inert. Geographically diversify your installations, population, business, industry, and military. Present decoy targets. And, when possible, do hunt down and destroy the enemy's capabilities.

I don't feel that the administration's handling is what it could be. If I were you, I'd be interested in finding out just what alert mechanisms (and communications methods) have been established in my neighborhood. NYC is both a likely target, and a well-defended zone. Rudy seems to be doing a pretty good job, considered (I wish he had someone else's job right now). And, be mindful that the enemy's attacks have maximized response, not effect.