Why does the assumption of independence of risks matter in the examples of insurance? What would happen to premiums if the probabilities of houses burning were positively correlated? Can you think of a situation where they might be negatively correlated?