There were 135[1] Space Shuttle missions with 2 resulting in human casualties (Challenger and Columbia disasters).
Thus, a failure with loss of vehicle and of human life of 1.48 in 100.
The estimates range from roughly 1 in 100 to 1 in 100,000. The higher figures come from the working engineers, and the very low figures from management.
The reality was even more dangerous than the engineers had predicted, and far more dangerous than management had.
One of the major problems with the shuttle is that the design of the Space Shuttle meant that a loss of mission pretty much also meant the loss of the crew as well. A 1 in 70 accident rate is not nearly as big a deal if there's a 90%+ chance of the crew surviving the accident. (I just pulled the 90% number out of the air -- I've never seen any actual estimates for the effectiveness of launch escape systems in the event of accidents. In practice the success rate has been 1/1.)
I would suspect the engineering estimate had one or at most two sig figs, which isn't bad compared to results. I'm sure the management estimate had six sig figures of course.
When NASA was comparing designs for their post-shuttle rocket recently, they literally used four significant figures in the risk estimates [1], and used this estimate to pick the design. It seems crazy to me to have four sig figs of reliability for designs that were basically at the PowerPoint stage.
NASA started with the requirement that their new rocket have less than 1 in 1000 odds of loss of crew (LOC). They concluded that using an existing Atlas V had LOC odds of 1 in 957 (unacceptable), while the paper design of putting a capsule on top of a Shuttle booster had LOC odds of 1 in 1918 (totally acceptable). They then quoted this 1,918 number in a lot of places to justify the program.
This rocket was the Ares-I [2], which turned into a fiasco and was canceled four years ago.
My conclusion is that NASA's current risk assessments are as bogus as the ones for the space shuttle. They start with an unrealistic goal (1 in 1000 risk), make totally unjustifiable estimates to meet the goal, and then make bad decisions based on these estimates. Coincidentally, the decisions based on these estimates line up with the politically-desirable outcome.
The 1 in 1918 risk assessment turned out to be totally wrong, of course. The Air Force pointed out that the launch escape system wouldn't work since burning fuel would melt the parachute and everyone would die. [3]
My personal view is that NASA needs to admit that rockets are dangerous and you probably can't get the risk below 1 in 100. Then NASA can focus on doing the best job they can. [4]
Thus, a failure with loss of vehicle and of human life of 1.48 in 100.
The estimates range from roughly 1 in 100 to 1 in 100,000. The higher figures come from the working engineers, and the very low figures from management.
The reality was even more dangerous than the engineers had predicted, and far more dangerous than management had.
[1] http://en.wikipedia.org/wiki/List_of_space_shuttle_missions