The book has two main arguments.
First, markets cannot price all forms of risk because there are some things we do not know (drumroll Donald Rumsfeld please)--what Skidelsky calls the "irreducible minimum". The models of the Neoclassical and, to a lesser extent, Neo-Keynsian economists however value "beauty over truth" (see Paul Krugman in this week's New York Times magazine). They want to cover all eventualities and the way they do that is to circumscribe all eventualities in a false set, defined by the models themselves. In this way the models don't even entertain the fact that some markets are not self correcting. They assume frictionless markets with cool calculating agents operating with perfect information. As Skidelsy notes, macroeconomics should be built on expectations that are human or conventional, not on those that are rational.
Second, economics should think of itself as a moral not a natural science. It should not be so dominated by market efficiency and the pursuit of money. Money is the "continual stimulator of our imagination creatign a perpetual sense of dissatisfaction".
Economics should have a "more modest role as tutor of governments"
- A new synthesis of government and market action
- Build financial models that can cope with irreducible uncertainty and make governments focus on reducing the impact of uncertainty
- Develop a new global reserve currency that is not influenced by any one country's needs