Adaptive Markets is A Random Walk Down Wall Street meets Thinking, Fast and Slow (with a healthy splash of The Selfish Gene).
Andrew Lo gives a fantastic overview of the origins of the Efficient Market Hypothesis and the difficulties of convincing academic finance that it is incorrect. However, I like that he doesn't go as far as completely discounting the value of the EMH – it is an incredible first approximation, and unlike so many other competing philosophies, it actually makes testable predictions (even if they turn out not to be correct).
A central theme in the book is that "it takes a theory to beat a theory" – the library of cognitive biases explored by Kahneman and Tversky does not quite meet this standard (even when unified under Prospect Theory). Lo argues that a prevailing issue is the modern economist's "box of tools" – post Samuelson, economists became closet wannabes of mathematical physicists. But it could just be that mathematical physics is fundamentally the wrong language to describe a complex system in which the fundamental unit does not obey deterministic rules – there's that famous Feynman quote about how much harder physics would be if electrons had feelings.
Lo instead proposes that the answer lies in evolutionary biology, which explains the origins of our biases, combined with the science of ecosystems which explains the complex feedback loops that process our individual behaviours into macroscopic properties. One of the highlights of the book is when Lo shows that this new theory is able to explain a particularly weird cognitive bias called probability matching.
The book is really a 5-star read, except for one deep criticism. It seems to me that Lo has slightly lower standards for his theory of Adaptive Markets than he does for the alternatives. In particular, my fear is that the Adaptive Markets hypothesis is one of those blanket theories that can predict anything under the sun – i.e it is not falsifiable in Popper's sense (though strangely, Lo uses that to criticise other theories). With the evolutionary simulations, for example, it is very clear that the parameters (which the researchers input) completely determine the outcome, and Lo neglects to discuss the difficult task of calibrating these agent-based models. I would have liked to see a slightly more self-critical approach to his own theory.
But this aside, Adaptive Markets is a refreshingly modern take on traditional finance and ends on a surprisingly optimistic note. After all, in nature, evolution has created wonderful diversity, efficiency, and beauty –why can't the same be true for financial markets?