How Should We Use Entropy in Economics? (1991) [pdf]

How Should We Use Entropy in Economics? (1991) [pdf](bayes.wustl.edu)

29 points by poindontcare 9 years ago | 6 comments

> Presumably, a process of trial and error will be needed to find the right choices

This is a huge difference between macroeconomics and thermodynamics. You can't just have a control group of people, test a policy with them and say 'omg they died of starvation'. I know well designed experiments exist such as basic income, but I hope I made a point: You can't just test macro policies as you would test heat transfer in a lab.

sit12 9 years ago | |

maybe you can model or simulate somehow?

ravar 9 years ago |

My first reaction is that there is no reason to use entropy. In physics you can prove that Hamiltonian motion leads to ergodic behavior and that creates the push to higher entropy. In an economic system I question that ergodicity would be satisfied. So similarly I doubt that there is inherent reason to expect entropy to increase or even be a useful metric on the system.

nerdponx 9 years ago | |

I think in a lot of economic systems we're still a long way off from reasoning at all about long run behavior, let alone formally deriving it. Hence the urge to try and steal modeling concepts from other fields, with or without theoretical foundations.

That said, I have very shaky intuition about how physicists actually use these concepts. Is egodicity necessary or just sufficient for rising entropy?

agentofoblivion 9 years ago | |

Although entropy originated in the context of statistical mechanics, Shannon/Jaynes showed it's a much more general and powerful concept. E.g. Information theory. Entropy is simply the average information in a probability distribution. Jaynes papers are super interesting! He derives famous stat mech results while making no ergodic assumptions and shows they're merely a consequence of statistical inference.

I suspect the concept is already present in economics. It's certainly under the hood of many statistical modeling techniques. For instance, some tree learning algorithms use entropy reduction measures to decide how to split the data for the tree. You can even derive the normal distribution by merely maximizing the entropy under the constraint of known variance.

cm127 9 years ago | |

I see it analogous with respect to limited resources and interests, but it gets completely distorted when debt can be created from nothing and used as currency. What's going to happen when debt exceeds all resources in the world? I guess that's the real reason for shooting into space: paying the world's insane debt.