Can you really measure and model risk?

I am not saying statistical inference is useless. As I said in my first post, for life insurance and the risk of a claim due to death is predictable. In fact if you have a large enough risk pool, the projection can be accurate to an impressive degree. On the other hand, the risk of mortgage default is a very different animal. With the benefit of hindsight I understand today that it has to do with trillions of easy credit flowing into this country, with China and Japan together lending us more than 1.6 trillion. The potent power of low interest rate created the asset bubble and the bust that follows. As you can see this is an external variable that keeps accumulating (by just look at the trade deficit). But AIG keeps selling Credit Default Swap (CDS) as if it is a life insurance to insure mortgage default, to the degree that without 80 billion of government bailout, it means chain bankruptcy of financial institutions all over the world and the Great Depression 2.0. I don't want to elaborate on the relationship between Great Depression 1.0 and World War 2 as I am not an expert, but you understand the scare. So let me conclude by answering to the question "Can you really measure and model risk?". My thought is that it depends on what kind of risk you are talking about. You need to use a common sense judgement before you apply the state-of-the-art software. And you need to have a sufficient understanding of the boom and bust nature of economic history. And when someone pretend to be able to model things that can't be on a sound scientific footing, you should be able to call their bluff and air your skepticism. It is healthy for quants to air their skepticism. You are helping the society by demystifying the quants.